OK, now it's crashing with an assertion.

[neroden]

Running my patched and compiled simutrans-experimental, crashes during map creation with this assertion:

simutrans-experimental: simfab.cc:289: fabrik_t::~fabrik_t(): Assertion `plan' failed.

(Why is it running a factory *destructor* during map creation?)

If I put in a settings.xml file it instead segfaults.

[VS]

Which "map creation" ? The one directly on startup before rolling credits screen, or the one you start from new world dialogue? In the latter case the old map must be removed

[neroden]

Which "map creation" ? The one directly on startup before rolling credits screen, or the one you start from new world dialogue? In the latter case the old map must be removed

Ah-HA.  It's the one you start from new world dialogue.  So it's probably dying dismantling the old map....

[neroden]

OK, I can generate a *different* crash-to-assertion bug.  

If I go into public service mode (on the initial map) and start deleting factories, eventually I'll get this crash:
sim: simfab.cc:295: fabrik_t::~fabrik_t(): Assertion `f' failed.

It's not completely predictable; I appear to be able to delete some before getting it.

Deleting the Walmington-on-Sea Forest/Sawmill seems to be the most reliable way of prodicing *this* crash.

The *GOOD* news is, if I delete the demo.sve file from Pak128.Britain.Ex 0.5.before starting up, then everything works.  (So far.)  So it's a save file compatibility issue.

And now I can actually test the game.   

[jamespetts]

Thank you very much indeed for your reports - much appreciated. The problem appears to be that it is attempting to delete fields that don't exist. In a version compiled with the -NDEBUG switch on, the assertion will simply be passed by, which is what should be the case for the release version. I presume that you compiled your own in Linux without the -NDEBUG switch?

[neroden]

Thank you very much indeed for your reports - much appreciated. The problem appears to be that it is attempting to delete fields that don't exist. In a version compiled with the -NDEBUG switch on, the assertion will simply be passed by, which is what should be the case for the release version. I presume that you compiled your own in Linux without the -NDEBUG switch?

Yes.  I had to do that to get a working version, as documented in other bug threads.