It's okay to have the window off the top/left edge of the primary
display, but CiderPress was rejecting the values and reverting to
default placement.
(for issue #41)
Windows' default behavior is apparently to fill the display with the
app window, capped at 1920x1200. This is annoyingly large for most
situations.
We now save the main window rect (LTRB) and maximization status in
the configuration area of the registry. The window placement calls
are supposed to do something reasonable when the window would be
completely off-screen (e.g. because a secondary monitor was disabled).
Windows is currently creating the main window at near-maximal size.
It used to remember the size and placement, but no longer does.
As a workaround, the initial size is set to 1150x800, which is large
enough to show all columns without scrolling even with very wide
pathnames. With some effort this could be modified to respect the
maximum size of the monitor on which it will be displayed, so that
anyone still running at 1024x768 won't be in a bad place.
Ideally it would remember the previous size and position. See
issue #41 for discussion.
DiskCopy disk images on HFS volumes, such as the ByteWorks Opus ][
CD-ROM, have resource forks. The double-click handler was screening
out forked files, so double-clicking on one of these disk images
was popping open the file viewer instead of creating a new instance
of CiderPress.
For a file with ".shk" extension, prioritize disk image over NuFX
file archive. The disk image code won't accept it if it has more
than one entry, and it's pretty rare for somebody to want to open
a single-disk archive in file mode.
Ideally all single-disk archives are named ".sdk", but there's
probably a lot that are just named ".shk", so this will probably
be more convenient.
The previous code was a stickler for only opening files whose type
matched what was selected in the filter pop-up. The original goal
was to allow you to choose whether a BXY or SDK file was interpreted
as Binary II, ShrinkIt, or disk image, since they could go either
way. Unfortunately, its refusal to consider types other than what
was selected made it kind of annoying.
The new code will start by trying to open the file with the selected
filter, so that it's still possible to choose how SDK and BXY files
are opened. However, it now continues on, trying all other types
before finally giving up.
If the generic ("*.*") filter is selected, CiderPress will start by
trying to open the file as a disk image.
This seems to produce good results with a variety of known and
unknown files.
This handles version 1 and 2, and copes with the broken files
created by the Mac OS X "applesingle" command-line tool (which is
unable to decode the broken files it creates).
I get the sense that many AppleSingle files don't end with ".AS", so
the filespec includes "*.*" as well.
Some AppleSingle files don't include a filename. In that case, we
use the file's name as the entry name, minus any ".as" extension.
The current implementation doesn't convert from Unicode to Mac OS
Roman, so non-ASCII characters are mishandled unless the file was
generated by GS/ShrinkIt. (We assume version 1 AppleSingle files
use MOR name strings.)
Also, version bump to 4.0.0d3.
This updates GenericEntry's filename handling to be more careful
about Mac OS Roman vs. Unicode. Most of the work is still done with
a CP-1252 conversion instead of MOR, but we now do a proper
conversion on the "display name", so we see the right thing in the
content list and file viewer.
Copy & paste, disk-to-file-archive, and file-archive-to-disk
conversions should work (correctly) as before. Extracted files will
still have "_" or "%AA" instead of a Unicode TRADE MARK SIGN, but
that's fine for now -- we can extract and re-add the files losslessly.
The filenames are now stored in CStrings rather than WCHAR*.
Also, fixed a bad initializer in the file-archive-to-disk conversion
dialog.
Most of this change is a conversion of the old FileDetails struct
into a new LocalFileDetails class. The new class keeps the
members private, and keeps the Unicode and MOR representations of
the string separate.
The NuFX and DiskImg libraries don't support UTF-16 filenames,
so we stil can't add files with non-CP-1252 filenames, but we're
a step closer.
Also, update NufxLib with a couple of fixes from the main project.
Also, fix handling of "%00" when adding files.
Also, mark most of the A2FileDOS fields private. Not sure why
they weren't.
DeployMaster can detect whether or not CiderPress is currently
running by checking for the presence of a window with a specific class
name. The default class name is generated differently each time, so
we need to set a custom class name.
Also, bumped version to 4.0.0d2.
The static analyzer was annoyed that the return value from calls to
CString::LoadString() was being ignored. This adds a wrapper
function that checks the value and logs a failure message if the
string can't be found.
The DeployMaster installer issue prevents the user from seeing more
than nine of the 18 file extensions that CiderPress wants to handle,
and I don't want to go stomping on file associations without some
way to disable the behavior. So this returns to the previous behavior,
where CiderPress directly manages the file associations.
The CiderPress app is not able to modify HKEY_LOCAL_MACHINE (which
it used to do via HKEY_CLASSES_ROOT) on recent versions of Windows --
tested in Win7, but it probably broke with Vista. So now we do
everything in HKEY_CURRENT_USER. This works, more or less.
We're not looking at the Windows shell overrides, which are made
in yet another set of registry entries, so there are multiple
reasons why the values reported by the Edit Associations dialog may
now be inaccurate. I still favor eliminating the dialog as a
long-term strategy.
I took the opportunity to do some code cleanup in the registry code.
I also added calls to SHChangeNotify() to tell the Windows shell when
file associations change, so Windows Explorer windows get updated
promptly.
In the past, CiderPress managed its own file associations. This is
the feature that launches CiderPress when you double-click on a ".shk"
file. The installer ran "CiderPress -install" and "-uninstall" during
installation and removal to give CP a chance to establish and clean
up the necessary registry entries.
The code built with VS6 works fine. The code built with VS2013 fails
with an access denied error. It appears there have been some access
policy changes, and the older code is getting "grandfathered in". This
is really something that the installer ought to be handling, though,
so rather than figure out how to fix CiderPress, I'm removing the
file type association code from CiderPress and letting DeployMaster
handle it.
This may be slightly less convenient for anyone who had reason to
change type associations frequently. Modern versions of Windows have
relatively easy to use control panel UIs for adjusting types, and
the "advanced installation" feature of DeployMaster allows you to
un-check the types that you don't want to have associated with
CiderPress.
(...with one minor hitch: DeployMaster 4.2.2 only shows the first 9
associations, and CiderPress has 18.)
This change renders most of the registry-handling code obsolete, as
well as the "-install" / "-uninstall" handling. I'm 99% sure I want
to go this way, but I'm keeping things #ifdefed rather than deleted
for the moment.
This moves method comments from the .cpp file to the .h file,
where users of the methods can find them. This also makes it
possible for the IDE to show the comments when you mouse-hover over
the method name, though Visual Studio is a bit weak in this regard.
Also, added "override" keywords on overridden methods. Reasonably
current versions of popular compilers seem to support this.
Also, don't have the return type on a separate line in the .cpp file.
The motivation for the practice -- quickly finding a method definition
with "^name" -- is less useful in C++ than C, and modern IDEs provide
more convenient ways to do the same thing.
Also, do some more conversion from unsigned types to uintXX_t.
This commit is primarily for the "app" directory.
Mostly a bulk conversion of debug messages, primarily with sed:
sed -e 's/\(WMSG[0-9]\)\(.*\)\(\\n"\)/LOGI\2"/'
This removes the '\n' from the end of the log messages, and sets
them all to "info" severity.
We want to prefix each line with file/line and/or a timestamp,
so it doesn't make sense to have a partial line, and there's no
value in embedding the '\n' in every string.
CiderPress and MDC now compile, and execute far enough to open
their respective "about" boxes, but I doubt they'll do much
more than that.
* Switch from MBCS to UNICODE APIs
Microsoft switched to UTF-16 (by way of UCS-2) a long time ago,
and the support for MBCS seems to be getting phased out. So it's
time to switch to wide strings.
This is a bit awkward for CiderPress because it works with disk
and file archives with 8-bit filenames, and I want NufxLib and
DiskImgLib to continue to work on Linux (which has largely taken
the UTF-8 approach to Unicode). The libraries will continue to
work with 8-bit filenames, with CiderPress/MDC doing the
conversion at the appropriate point.
There were a couple of places where strings from a structure
handed back by one of the libraries were used directly in the UI,
or vice-versa, which is a problem because we have nowhere to
store the result of the conversion. These currently have fixed
place-holder "xyzzy" strings.
All UI strings are now wide.
Various format strings now use "%ls" and "%hs" to explicitly
specify wide and narrow. This doesn't play well with gcc, so
only the Windows-specific parts use those.
* Various updates to vcxproj files
The project-file conversion had some cruft that is now largely
gone. The build now has a common output directory for the EXEs
and libraries, avoiding the old post-build copy steps.
* Added zlib 1.2.8 and nufxlib 2.2.2 source snapshots
The old "prebuilts" directory is now gone. The libraries are now
built as part of building the apps.
I added a minimal set of files for zlib, and a full set for nufxlib.
The Linux-specific nufxlib goodies are included for the benefit of
the Linux utilities, which are currently broken (don't build).
* Replace symbols used for include guards
Symbols with a leading "__" are reserved.
This updates all source files to use spaces instead of tabs for
indentation. It also normalizes the end-of-line markers to be
Windows-style CRLF, and ensures that all files end with EOL.
No substantive changes were made; "diff -w" is empty.
Andy McFadden
Andy McFadden
David Schmidt