2014-06-15 19:34:26 +00:00
|
|
|
; REQUIRES: object-emission
|
|
|
|
|
|
|
|
; RUN: %llc_dwarf -O0 -filetype=obj < %s | llvm-dwarfdump -debug-dump=info - | FileCheck %s
|
|
|
|
|
|
|
|
; IR generated from clang -O0 with:
|
|
|
|
; struct C {
|
|
|
|
; ~C();
|
|
|
|
; };
|
|
|
|
; extern bool b;
|
|
|
|
; void fun4() { b && (C(), 1); }
|
|
|
|
; __attribute__((always_inline)) C::~C() { }
|
|
|
|
|
|
|
|
; CHECK: DW_TAG_structure_type
|
|
|
|
; CHECK-NOT: DW_TAG
|
|
|
|
; CHECK: DW_AT_name {{.*}} "C"
|
|
|
|
; CHECK-NOT: {{DW_TAG|NULL}}
|
|
|
|
; CHECK: [[C_DTOR_DECL:.*]]: DW_TAG_subprogram
|
|
|
|
; CHECK-NOT: DW_TAG
|
|
|
|
; CHECK: DW_AT_name {{.*}} "~C"
|
|
|
|
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
; CHECK: [[D1_ABS:.*]]: DW_TAG_subprogram
|
2014-06-27 05:31:49 +00:00
|
|
|
; CHECK-NOT: DW_TAG
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
; CHECK: DW_AT_MIPS_linkage_name {{.*}} "_ZN1CD1Ev"
|
|
|
|
; CHECK-NOT: {{DW_TAG|NULL}}
|
|
|
|
; CHECK: [[D1_THIS_ABS:.*]]: DW_TAG_formal_parameter
|
|
|
|
; CHECK-NOT: DW_TAG
|
|
|
|
; CHECK: DW_AT_name {{.*}} "this"
|
2014-07-01 04:11:45 +00:00
|
|
|
|
|
|
|
; CHECK: DW_TAG_subprogram
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
and I now believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212085 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-01 03:11:59 +00:00
|
|
|
; CHECK-NOT: DW_TAG
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
; CHECK: DW_AT_name {{.*}} "fun4"
|
|
|
|
; CHECK-NOT: {{DW_TAG|NULL}}
|
2014-08-31 21:26:22 +00:00
|
|
|
; CHECK: DW_TAG_inlined_subroutine
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
; CHECK-NOT: DW_TAG
|
2014-08-31 21:26:22 +00:00
|
|
|
; CHECK: DW_AT_abstract_origin {{.*}} {[[D1_ABS]]}
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
; CHECK-NOT: {{DW_TAG|NULL}}
|
2014-08-31 21:26:22 +00:00
|
|
|
; CHECK: DW_TAG_formal_parameter
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
; CHECK-NOT: DW_TAG
|
2014-08-31 21:26:22 +00:00
|
|
|
; CHECK: DW_AT_abstract_origin {{.*}} {[[D1_THIS_ABS]]}
|
2014-07-01 04:11:45 +00:00
|
|
|
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
; FIXME: D2 is actually inlined into D1 but doesn't show up here, possibly due
|
|
|
|
; to there being no work in D2 (calling another member function from the dtor
|
|
|
|
; causes D2 to show up, calling a free function doesn't).
|
2014-07-01 04:11:45 +00:00
|
|
|
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
; CHECK-NOT: DW_TAG
|
|
|
|
; CHECK: NULL
|
|
|
|
; CHECK-NOT: DW_TAG
|
|
|
|
; CHECK: NULL
|
2014-06-15 19:34:26 +00:00
|
|
|
|
|
|
|
%struct.C = type { i8 }
|
|
|
|
|
|
|
|
@b = external global i8
|
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
|
|
|
define void @_Z4fun4v() #0 {
|
|
|
|
entry:
|
|
|
|
%this.addr.i.i = alloca %struct.C*, align 8, !dbg !21
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
%this.addr.i = alloca %struct.C*, align 8, !dbg !22
|
2014-06-15 19:34:26 +00:00
|
|
|
%agg.tmp.ensured = alloca %struct.C, align 1
|
|
|
|
%cleanup.cond = alloca i1
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
%0 = load i8* @b, align 1, !dbg !24
|
|
|
|
%tobool = trunc i8 %0 to i1, !dbg !24
|
2014-06-15 19:34:26 +00:00
|
|
|
store i1 false, i1* %cleanup.cond
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
br i1 %tobool, label %land.rhs, label %land.end, !dbg !24
|
2014-06-15 19:34:26 +00:00
|
|
|
|
|
|
|
land.rhs: ; preds = %entry
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
store i1 true, i1* %cleanup.cond, !dbg !25
|
2014-06-15 19:34:26 +00:00
|
|
|
br label %land.end
|
|
|
|
|
|
|
|
land.end: ; preds = %land.rhs, %entry
|
|
|
|
%1 = phi i1 [ false, %entry ], [ true, %land.rhs ]
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
%cleanup.is_active = load i1* %cleanup.cond, !dbg !27
|
|
|
|
br i1 %cleanup.is_active, label %cleanup.action, label %cleanup.done, !dbg !27
|
2014-06-15 19:34:26 +00:00
|
|
|
|
|
|
|
cleanup.action: ; preds = %land.end
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
store %struct.C* %agg.tmp.ensured, %struct.C** %this.addr.i, align 8, !dbg !22
|
2014-10-03 20:01:09 +00:00
|
|
|
call void @llvm.dbg.declare(metadata !{%struct.C** %this.addr.i}, metadata !29, metadata !{metadata !"0x102"}), !dbg !31
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
%this1.i = load %struct.C** %this.addr.i, !dbg !22
|
2014-06-15 19:34:26 +00:00
|
|
|
store %struct.C* %this1.i, %struct.C** %this.addr.i.i, align 8, !dbg !21
|
2014-10-03 20:01:09 +00:00
|
|
|
call void @llvm.dbg.declare(metadata !{%struct.C** %this.addr.i.i}, metadata !32, metadata !{metadata !"0x102"}), !dbg !33
|
2014-06-15 19:34:26 +00:00
|
|
|
%this1.i.i = load %struct.C** %this.addr.i.i, !dbg !21
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
br label %cleanup.done, !dbg !22
|
2014-06-15 19:34:26 +00:00
|
|
|
|
|
|
|
cleanup.done: ; preds = %cleanup.action, %land.end
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
ret void, !dbg !34
|
2014-06-15 19:34:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
; Function Attrs: alwaysinline nounwind
|
|
|
|
define void @_ZN1CD1Ev(%struct.C* %this) unnamed_addr #1 align 2 {
|
|
|
|
entry:
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
%this.addr.i = alloca %struct.C*, align 8, !dbg !37
|
2014-06-15 19:34:26 +00:00
|
|
|
%this.addr = alloca %struct.C*, align 8
|
|
|
|
store %struct.C* %this, %struct.C** %this.addr, align 8
|
2014-10-03 20:01:09 +00:00
|
|
|
call void @llvm.dbg.declare(metadata !{%struct.C** %this.addr}, metadata !29, metadata !{metadata !"0x102"}), !dbg !38
|
2014-06-15 19:34:26 +00:00
|
|
|
%this1 = load %struct.C** %this.addr
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
store %struct.C* %this1, %struct.C** %this.addr.i, align 8, !dbg !37
|
2014-10-03 20:01:09 +00:00
|
|
|
call void @llvm.dbg.declare(metadata !{%struct.C** %this.addr.i}, metadata !32, metadata !{metadata !"0x102"}), !dbg !39
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
%this1.i = load %struct.C** %this.addr.i, !dbg !37
|
|
|
|
ret void, !dbg !37
|
2014-06-15 19:34:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
; Function Attrs: alwaysinline nounwind
|
|
|
|
define void @_ZN1CD2Ev(%struct.C* %this) unnamed_addr #1 align 2 {
|
|
|
|
entry:
|
|
|
|
%this.addr = alloca %struct.C*, align 8
|
|
|
|
store %struct.C* %this, %struct.C** %this.addr, align 8
|
2014-10-03 20:01:09 +00:00
|
|
|
call void @llvm.dbg.declare(metadata !{%struct.C** %this.addr}, metadata !32, metadata !{metadata !"0x102"}), !dbg !40
|
2014-06-15 19:34:26 +00:00
|
|
|
%this1 = load %struct.C** %this.addr
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
ret void, !dbg !41
|
2014-06-15 19:34:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
; Function Attrs: nounwind readnone
|
Move the complex address expression out of DIVariable and into an extra
argument of the llvm.dbg.declare/llvm.dbg.value intrinsics.
Previously, DIVariable was a variable-length field that has an optional
reference to a Metadata array consisting of a variable number of
complex address expressions. In the case of OpPiece expressions this is
wasting a lot of storage in IR, because when an aggregate type is, e.g.,
SROA'd into all of its n individual members, the IR will contain n copies
of the DIVariable, all alike, only differing in the complex address
reference at the end.
By making the complex address into an extra argument of the
dbg.value/dbg.declare intrinsics, all of the pieces can reference the
same variable and the complex address expressions can be uniqued across
the CU, too.
Down the road, this will allow us to move other flags, such as
"indirection" out of the DIVariable, too.
The new intrinsics look like this:
declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr)
declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr)
This patch adds a new LLVM-local tag to DIExpressions, so we can detect
and pretty-print DIExpression metadata nodes.
What this patch doesn't do:
This patch does not touch the "Indirect" field in DIVariable; but moving
that into the expression would be a natural next step.
http://reviews.llvm.org/D4919
rdar://problem/17994491
Thanks to dblaikie and dexonsmith for reviewing this patch!
Note: I accidentally committed a bogus older version of this patch previously.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
|
|
|
declare void @llvm.dbg.declare(metadata, metadata, metadata) #2
|
2014-06-15 19:34:26 +00:00
|
|
|
|
|
|
|
attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-realign-stack" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
attributes #1 = { alwaysinline nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-realign-stack" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
attributes #2 = { nounwind readnone }
|
|
|
|
|
|
|
|
!llvm.dbg.cu = !{!0}
|
|
|
|
!llvm.module.flags = !{!18, !19}
|
|
|
|
!llvm.ident = !{!20}
|
|
|
|
|
2014-10-03 20:01:09 +00:00
|
|
|
!0 = metadata !{metadata !"0x11\004\00clang version 3.5.0 \000\00\000\00\001", metadata !1, metadata !2, metadata !3, metadata !11, metadata !2, metadata !2} ; [ DW_TAG_compile_unit ] [/tmp/dbginfo/<stdin>] [DW_LANG_C_plus_plus]
|
2014-06-15 19:34:26 +00:00
|
|
|
!1 = metadata !{metadata !"<stdin>", metadata !"/tmp/dbginfo"}
|
|
|
|
!2 = metadata !{}
|
|
|
|
!3 = metadata !{metadata !4}
|
2014-10-03 20:01:09 +00:00
|
|
|
!4 = metadata !{metadata !"0x13\00C\001\008\008\000\000\000", metadata !5, null, null, metadata !6, null, null, metadata !"_ZTS1C"} ; [ DW_TAG_structure_type ] [C] [line 1, size 8, align 8, offset 0] [def] [from ]
|
2014-06-15 19:34:26 +00:00
|
|
|
!5 = metadata !{metadata !"PR20038.cpp", metadata !"/tmp/dbginfo"}
|
|
|
|
!6 = metadata !{metadata !7}
|
2014-10-03 20:01:09 +00:00
|
|
|
!7 = metadata !{metadata !"0x2e\00~C\00~C\00\002\000\000\000\006\00256\000\002", metadata !5, metadata !"_ZTS1C", metadata !8, null, null, null, i32 0, null} ; [ DW_TAG_subprogram ] [line 2] [~C]
|
|
|
|
!8 = metadata !{metadata !"0x15\00\000\000\000\000\000\000", i32 0, null, null, metadata !9, null, null, null} ; [ DW_TAG_subroutine_type ] [line 0, size 0, align 0, offset 0] [from ]
|
2014-06-15 19:34:26 +00:00
|
|
|
!9 = metadata !{null, metadata !10}
|
2014-10-03 20:01:09 +00:00
|
|
|
!10 = metadata !{metadata !"0xf\00\000\0064\0064\000\001088", null, null, metadata !"_ZTS1C"} ; [ DW_TAG_pointer_type ] [line 0, size 64, align 64, offset 0] [artificial] [from _ZTS1C]
|
2014-06-15 19:34:26 +00:00
|
|
|
!11 = metadata !{metadata !12, metadata !16, metadata !17}
|
2014-10-03 20:01:09 +00:00
|
|
|
!12 = metadata !{metadata !"0x2e\00fun4\00fun4\00_Z4fun4v\005\000\001\000\006\00256\000\005", metadata !5, metadata !13, metadata !14, null, void ()* @_Z4fun4v, null, null, metadata !2} ; [ DW_TAG_subprogram ] [line 5] [def] [fun4]
|
|
|
|
!13 = metadata !{metadata !"0x29", metadata !5} ; [ DW_TAG_file_type ] [/tmp/dbginfo/PR20038.cpp]
|
|
|
|
!14 = metadata !{metadata !"0x15\00\000\000\000\000\000\000", i32 0, null, null, metadata !15, null, null, null} ; [ DW_TAG_subroutine_type ] [line 0, size 0, align 0, offset 0] [from ]
|
2014-06-15 19:34:26 +00:00
|
|
|
!15 = metadata !{null}
|
2014-10-03 20:01:09 +00:00
|
|
|
!16 = metadata !{metadata !"0x2e\00~C\00~C\00_ZN1CD2Ev\006\000\001\000\006\00256\000\006", metadata !5, metadata !"_ZTS1C", metadata !8, null, void (%struct.C*)* @_ZN1CD2Ev, null, metadata !7, metadata !2} ; [ DW_TAG_subprogram ] [line 6] [def] [~C]
|
|
|
|
!17 = metadata !{metadata !"0x2e\00~C\00~C\00_ZN1CD1Ev\006\000\001\000\006\00256\000\006", metadata !5, metadata !"_ZTS1C", metadata !8, null, void (%struct.C*)* @_ZN1CD1Ev, null, metadata !7, metadata !2} ; [ DW_TAG_subprogram ] [line 6] [def] [~C]
|
2014-06-15 19:34:26 +00:00
|
|
|
!18 = metadata !{i32 2, metadata !"Dwarf Version", i32 4}
|
2014-10-03 20:01:09 +00:00
|
|
|
!19 = metadata !{i32 2, metadata !"Debug Info Version", i32 2}
|
2014-06-15 19:34:26 +00:00
|
|
|
!20 = metadata !{metadata !"clang version 3.5.0 "}
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
!21 = metadata !{i32 6, i32 0, metadata !17, metadata !22}
|
|
|
|
!22 = metadata !{i32 5, i32 0, metadata !23, null}
|
2014-10-03 20:01:09 +00:00
|
|
|
!23 = metadata !{metadata !"0xb\005\000\003", metadata !5, metadata !12} ; [ DW_TAG_lexical_block ] [/tmp/dbginfo/PR20038.cpp]
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
!24 = metadata !{i32 5, i32 0, metadata !12, null}
|
|
|
|
!25 = metadata !{i32 5, i32 0, metadata !26, null}
|
2014-10-03 20:01:09 +00:00
|
|
|
!26 = metadata !{metadata !"0xb\005\000\001", metadata !5, metadata !12} ; [ DW_TAG_lexical_block ] [/tmp/dbginfo/PR20038.cpp]
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
!27 = metadata !{i32 5, i32 0, metadata !28, null}
|
2014-10-03 20:01:09 +00:00
|
|
|
!28 = metadata !{metadata !"0xb\005\000\002", metadata !5, metadata !12} ; [ DW_TAG_lexical_block ] [/tmp/dbginfo/PR20038.cpp]
|
|
|
|
!29 = metadata !{metadata !"0x101\00this\0016777216\001088", metadata !17, null, metadata !30} ; [ DW_TAG_arg_variable ] [this] [line 0]
|
|
|
|
!30 = metadata !{metadata !"0xf\00\000\0064\0064\000\000", null, null, metadata !"_ZTS1C"} ; [ DW_TAG_pointer_type ] [line 0, size 64, align 64, offset 0] [from _ZTS1C]
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
!31 = metadata !{i32 0, i32 0, metadata !17, metadata !22}
|
2014-10-03 20:01:09 +00:00
|
|
|
!32 = metadata !{metadata !"0x101\00this\0016777216\001088", metadata !16, null, metadata !30} ; [ DW_TAG_arg_variable ] [this] [line 0]
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
!33 = metadata !{i32 0, i32 0, metadata !16, metadata !21}
|
|
|
|
!34 = metadata !{i32 5, i32 0, metadata !35, null}
|
2014-10-03 20:01:09 +00:00
|
|
|
!35 = metadata !{metadata !"0xb\005\000\005", metadata !5, metadata !36} ; [ DW_TAG_lexical_block ] [/tmp/dbginfo/PR20038.cpp]
|
|
|
|
!36 = metadata !{metadata !"0xb\005\000\004", metadata !5, metadata !12} ; [ DW_TAG_lexical_block ] [/tmp/dbginfo/PR20038.cpp]
|
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself.
Originally committed in r211723, reverted in r211724 due to failure
cases found and fixed (ArgumentPromotion: r211872, Inlining: r212065),
committed again in r212085 and reverted again in r212089 after fixing
some other cases, such as debug info subprogram lists not keeping track
of the function they represent (r212128) and then short-circuiting
things like LiveDebugVariables that build LexicalScopes for functions
that might not have full debug info.
And again, I believe the invariant actually holds for some reasonable
amount of code (but I'll keep an eye on the buildbots and see what
happens... ).
Original commit message:
PR20038: DebugInfo: Inlined call sites where the caller has debug info
but the call itself has no debug location.
This situation does bad things when inlined, so I've fixed Clang not to
produce inlinable call sites without locations when the caller has debug
info (in the one case where I could find that this occurred). This
updates the PR20038 test case to be what clang now produces, and readds
the assertion that had to be removed due to this bug.
I've also beefed up the debug info verifier to help diagnose these
issues in the future, and I hope to add checks to the inliner to just
assert-fail if it encounters this situation. If, in the future, we
decide we have to cope with this situation, the right thing to do is
probably to just remove all the DebugLocs from the inlined instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@212205 91177308-0d34-0410-b5e6-96231b3b80d8
2014-07-02 18:32:05 +00:00
|
|
|
!37 = metadata !{i32 6, i32 0, metadata !17, null}
|
|
|
|
!38 = metadata !{i32 0, i32 0, metadata !17, null}
|
|
|
|
!39 = metadata !{i32 0, i32 0, metadata !16, metadata !37}
|
|
|
|
!40 = metadata !{i32 0, i32 0, metadata !16, null}
|
|
|
|
!41 = metadata !{i32 6, i32 0, metadata !16, null}
|