The bug is that an #if condition may be mis-evaluated if it immediately follows a use of a function-like macro. I also modified a couple other places that could possibly trigger the bug (including a possibly more reliable fix for the instance of this problem I worked around earlier in include/xatonum.h).
We do this by maintaining a mask (for each pid) giving the fds to be closed on exec. We wrap functions that close fds so that their close-on-exec bits can be cleared at that point.
This implementation may close the fds even if the execve operation ultimately fails.
*Most significantly, we avoid using setpgid(), because it doesn't work and in fact corrupts the kernel's process group table.
*Also, work around tctpgrp() returning garbage instead of 0 on success.
This adds an implementation of tcsetpgrp that works by reading the process tables to find a process in the appropriate group. This isn't used for the main job control operations, though, since it might be relatively slow.
At this point, basic job control seems to work.
* Push/pop environment to make sure it is isolated from our parents and children.
* Make all environment vars (and shell vars) case-insensitive, consistent with GNO's internal handling of environment vars.
* Wrap putenv and unsetenv to make sure they are called with lower-case variable names, which is necessary to maintain consistency between the environ array and the kernel's internal representation of variables.
This should avoid strange behavior due to races when the parent has resumed but the child is still running the exec* code in libc, which mainly manifests itself when running at low speed.
We also change to signaling the child's completion with SIGALRM, and setting an extra alarm in the parent in case the child doesn't actually do it.
GNO's _exit (contrary to its man page) does clean-up for stuff like the memory allocator, which is inappropriate in a forked child process and leads to hangs and crashes.
*Use .null instead of /dev/null
*Account for GNO's dup2(), which non-standardly returns 0 on success
*Always call open with appropriate number of arguments
*Use STDIN_FILENO instead of (implicitly) 0
In particular, it will buffer information about children other than the one being waited for, so it isn't lost. It can also emulate a non-blocking wait by arranging to interrupt the wait with a signal.
This involved breaking things up into more segments in debug mode, since the code is larger. I also had to remove some unused extern definitions, which were causing link errors when debug code was enabled.
To enable debug code, pass "DEBUG=1" to make or build.gs.
This involved using signal() rather than sigaction(), which GNO doesn't have. signal() doesn't have the signal-blocking feature of sigaction(), but I've tried to emulate it.
This disables long argument support, since that requires getopt_long(), which isn't included in GNO. There is a separate library available for it, but it's simpler to avoid depending on that.
This is a BSD-licensed poll() implementation by Brian M. Clapper. Its performance characteristics aren't as good as a "native" poll() would be, but this shouldn't be a problem in practice.
Dmake gets confused if you use over-15-character file names, so they needed to be shortened to get it to work. While I was at it, I also switched _ to . in the filenames so they're fully ProDOS compatible.
The shell/hush_test directory (not needed to build the code) still uses ProDOS-incompatible file names.
This makes "G" (goto end of input) command work as well as
/search_for_nonexistent_string: both will read to EOF now
even from somewhat slow input (such as kernel's "git log").
function old new delta
ndelay_on 35 43 +8
ndelay_off 35 43 +8
read_lines 695 691 -4
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/1 up/down: 16/-4) Total: 12 bytes
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
We don't have an INIT_FIRST, so let's rename INIT_LAST to INIT_FUNC
to imply that the function is called at program start-up.
Also: the priority argument for __attribute__((constructor)) isn't
used, so let's remove it.
Signed-off-by: Bartosz Golaszewski <bartekgola@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This set of patches adds a simple unit-testing framework to Busybox
unit-tests: add some helper macros for unit-test framework implementation
unit-tests: implement the unit-testing framework
unit-tests: add basic documentation on writing the unit test cases
unit-tests: modify the Makefile 'test' target to run unit-tests too
unit-tests: add two example test cases
unit-tests: modify the existing strrstr test code to use the unit-test framework
Signed-off-by: Bartosz Golaszewski <bartekgola@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
The find utility uses a hardcoded value of 32 * 1024 as the limit of
the command-line length when calling 'find -exec ... {} +'. This results
in over 4 times more execve() calls than in coreutils' find.
This patch uses the limit defined in system headers.
Based on the patch by Bartosz Golaszewski <bartekgola@gmail.com>.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
