Bug#476217: psi vs Qt4.4.0-3 segmentation fault
On Monday 07 July 2008 10:40:05 Jan Niehusmann wrote:
> Hi Sune,
> (Cc: to all the other addresses, because just removing the 3 patches may
> not give meaningfull results.)
Removing those for now.
> I just did a similar thing: I did not comment the three patches, but
> just added a single if(isnan(...)) abort(); for debugging purposes. But
> now, the segfault doesn't occur any more.
> AFAICT, there are only two possible explanations: Either the bug doesn't
> happen with my toolchain (perhaps the compiler version is different than
> the one used to compile the debian qt package?), or it's some
> missoptimization by gcc which went away with the additional code line.
> Next, I'll try it with an unmodified source.
> If this still 'fixes' the problem, the core of the bug is much more
> difficult to find. The good news would be, that a simple recompile
> of qt on an up-to-date system probably would solve (or work around)
> the segfault.
I have been reviewing build logs between -2 and -3. and there isn't many
differences, except 1.
-2 is built with gcc4.2. -3 is built with gcc4.3.
So it might very well be a miscompilation or misoptimization. Those are though
a bit out of my scope of knowledge.
There has been a few gcc-4.3 uploads since then though.
Man, do you know how can I do for linking a server over the mailer?
From the file inside Internet Explorer you neither should ever ping to the
printer, nor have to send to a RO 3D periferic for mounting the coaxial SCSI