I have recently discovered two bugs in the CVS version of GIMP. The other
I have submitted as a bug report, but this one I really tried hard to fix
myself first, so I decided to send it along as a mail rather than attempt
to squeeze all this stuff into a web-report form.
GIMP does not save images sometimes, even if the dirty flag is set. This
one has been reported in the past, I believe -- it has been attributed to
the "trust_dirty_flag" option being set, or improperly checked somehow (at
least according to those who helped me in the #gimp(gimpnet) and #e(efnet)
IRC channels). I don't believe this is the case.
After some hunting around, I discovered that file_save was actually being
called. I inserted some debugging information in fileops.c, which I have
attached a diff of (attachment #3). This is against the most recent
anonymous CVS as of the writing of this email.
Although this problem appears to occurr on all fileformats, the only one I
can reliably repeat it on is .xpm. nosave.xpm (attachment #1) is a file
from an IceWM theme i was working on which, once opened, no matter how it
is altered, will not save. I have tried this 30 or 40 times, so I
consider it repeatable. :-)
I have also attached the output of a session where I open GIMP, change the
image (fill it with white, to be exact) then attempt to save it, then
quit, with the additional logging stuff in fileops.c.
The procedural_db_execute call appears to be failing, and I didn't have
enough time to start hunting through the procedural db or XPM code yet.
I figure that this will be enough of a head-start for a someone who
actually knows their way around the code already :-)
In case it matters: this is a RedHat 6.1 TOTALLY stock install, with one
very small exception (I add a TrueType fontserver (xfstt) to the stock xfs
that RH provides) . I wiped my harddrive and installed a "GNOME
Workstation" install before doing this test.
This means: gtk+/glib-1.2.5, glibc-2.1.2, egcs-2.91.66, XFree86-3.3.5.
----
The Tao is like a glob pattern: It is masked but always present.
used but never used up. I don't know who built to it.
It is like the extern void: It came before the first kernel.
filled with infinite possibilities. [glyph@xxxxxxxxxxxxxxxxx]
/* XPM */
static char * titleAL_xpm[] = {
"16 16 2 1",
" g None",
". g #000000",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................",
"................"};
Message: Passed serialization test
gimage is an active drawable
didn't find file proc first time
found file proc second time, continuing...
procedure: (file_xpm_save)
I conclude, returning 0.
Index: fileops.c
===================================================================
RCS file: /cvs/gnome/gimp/app/fileops.c,v
retrieving revision 1.77
diff -r1.77 fileops.c
987c987,994
< return FALSE;
---
> {
> g_print("gimage is not an active drawable\n");
> return FALSE;
> }
> else
> {
> g_print("gimage is an active drawable\n");
> }
992c999,1006
< file_proc = file_proc_find (save_procs, raw_filename);
---
> {
> g_print("didn't find file proc first time\n");
> file_proc = file_proc_find (save_procs, raw_filename);
> }
> else
> {
> g_print("found file proc first time, continuing...\n");
> }
995c1009,1016
< return FALSE;
---
> {
> g_print("didn't find file proc second time, giving up\n");
> return FALSE;
> }
> else
> {
> g_print("found file proc second time, continuing...\n");
> }
1010c1031
<
---
> g_print("procedure: (%s) \n",proc->name);
1049c1070
<
---
> g_print("I conclude, returning %d.\n",return_val);
