OK, I understand a bit more how the log works now (I never really got into Git after GZDoom moved from SVN to Git because it was at the time when there was becoming less and less need for me to be compiling the source). If I'm reading it right:
g4.7.0pre-101-g39513cf7a (
which works) was compiled after 39513cf7aef33dd0de3afd662815c64aafa870b5 which was committed on 9th August.
g4.7.0pre-131-gc3772fe20 (
which crashes) was compiled after c3772fe2036a8778d2e7198570a3ab0658e2e88e which was the last commit on 11th August.
Therefore, all commits from 10th August (13 of them) and 11th August (17 of them) comprise the difference between the working and the crashing build.
Is that correct?
That's a lot of commits (it would have to happen after two busy days with no builds in the middle) and none of the commits really leap into my non-coder brain as obvious "yes, that's the one" from their descriptions (though a few seem as if they are unlikely).
Unfortunately, my machine isn't set up for compiling anything at the moment so I can't even try to pick through them one by one and track it down any further than the above broad list myself, otherwise I'd have a go.
[edit]
I have also been able to get the crash with the Siren TC ( viewtopic.php?f=19&t=68686 ). The PDA messages in that are less graphically intensive than the ones in my file but after about 90 seconds of clicking around the PDA interface (after summoning all of the PDAs) I got the same type of crash.
It is perhaps interesting that the crash took longer to get in Sirens versus my test file. My messages use a lot of graphics and the Sirens ones are just text (plus, mine uses hi-res graphics and has more PDAs too) but they do both show the same kind of crash eventually.
Unfortunately, Sirens is also a huge project and, obviously, is also based on the PDA starter kit too. So slimming it down to a smaller test file is not really possible there either.
[/edit]