Startup crash (perhaps related to modified CFLAGS)?
Hi all,
for a while now I have been compiling Inkscape with empty CFLAGS/CXXFLAGS, which greatly speeds up the build. Since last week, however, Inkscape consistently crashes for me during startup. I first thought this was caused by one of the recent SVN commits, but bisecting the history I was unable to identify any malicious commit because a single revision sometimes worked and sometimes didn't. It's still strange that the problem only seems to occur with revisions that are not older than one or two weeks.
I have a separate SVN checkout on my hard drive which uses unchanged CFLAGS/CXXFLAGS (which I use when I need full debugging information). This one works without problems.
Has anyone else seen similar crashes in the past, perhaps related to modified CFLAGS? I'd be very glad if someone could point me to a solution because development is severely slowed down with the increased compile times, and I've already spent several days just trying to recompile Inksacpe and finding that the crash is still there.
Many thanks in advance, Max
P.S.: Of course I tried 'make clean' but that didn't help. :(( Backtraces are worthless, too, since they read like this:
#0 0xb74db9d7 in ?? () from /usr/lib/libgdk-x11-2.0.so.0 #1 0x00000000 in ?? ()
(as I said, debugging symbols are stripped off in these builds)
Hi,
a quick update from my side. After endless attempts, I think I have been able to narrow down the cause of my constant startup crashes. They definitely have something to do with the cairoification changes between rev #19024 and #19033. Unfortunately, the revisions in between don't compile as they are, which makes bisecting the history harder, but after making slight adaptions to recover compiling and a lot of experiments, I'm pretty sure that the following call in line #1685 of display/sp-canvas.cpp is the one to blame:
cairo_surface_t *cst = cairo_get_target(buf.ct);
When I comment it out (as well as the other two lines referring to the variable cst), Inkscape starts up as usual and I can work normally, except that rendering is of course slightly broken.
Since I'm not familiar with cairo, does anyone have a clue why this call could be causing a crash? I upgraded my cairo version from 1.5.12 to 1.6.4, just in case this was a bug on cairo's side, but it didn't help.:(
Do you think cairo is to blame and should I report a bug there? Or is cairo_get_target() simply passed invalid data so that this is fixable on Inkscape's side? Any help is greatly appreciated.
I still don't understand why the crashes only happen in some of my SVN checkouts, but perhaps my first guess about the modified CFLAGS was wrong and they don't have anything to do with the problem at all. Then again why don't more people see this?
Max
Maximilian Albert wrote:
I'm pretty sure that the following call in line #1685 of display/sp-canvas.cpp is the one to blame:
cairo_surface_t *cst = cairo_get_target(buf.ct);When I comment it out (as well as the other two lines referring to the variable cst), Inkscape starts up as usual and I can work normally, except that rendering is of course slightly broken.
Second update: I did these experiments on the revision where the mentioned call was introduced. When I comment out the line in the latest SVN version, rendering seems to be just fine (or if there are any artefacts they are at least not visible). So is this function call really needed? What does it do exactly?
Max
participants (1)
-
Maximilian Albert