Hi Przemek,
Unfortunately I cannot reproduce the problem myself. It may be
caused by different order of startup code execution in compilers
you are using.
If possible please try to use CodeGuard in BCC build. It should
catch the place where this code fails.
Also if possible please check if the problem is repeatable with
Harbour compiled without -gc2.
I have no CodeGuard unfortunately. I didn't debug Harbour
since 8 years, but I'll try to find something out.
ERR means GPF, plus when Harbour rebuilt with FM STATS ON,
it has shown this line too:
'Called from (_INITSTATICS00013)(0) in src\getsys.prg'
This is not Harbour code. We have only one file source\rtl\getsys.prg
which does not use any static variables. It will be funny if above
error comes from some old object files compiled before:
2008-10-13 12:42 UTC+0200 Przemyslaw Czerpak (druzus/at/
priv.onet.pl)
with broken static offsets.
Nope sorry, I left my framework lib in the liblist that's
all, and I have getsys.prg overloaded (and BCC build of the
framework weren't rebuilt since the problem). I didn't notice,
even if the problem your mention crossed my mind when seeing
the error. I've even double checked getsys.c, but the wrong
one. Sigh. :( This explains the funny message. Nevertheless,
it wasn't related to the main problem:
In the meantime I've rebuilt BCC 5.8.2 with plain default
build settings from a fresh SVN checkout. This clears the
HB_NO_DEBUG and -gc3 concerns, also possible local source
corruption. I also did a full default build of MSVC2008 from
fresh checkout, it didn't change the results.
Now results are fully consistent with MSVC, so that it also
shows the RTE and than GPFs (instead of giving the FM stats report),
in the same situations. That is with MT and -b and _no_ -l
switch. When -l switch is used, the problem goes away.
Brgds,
Viktor
_______________________________________________
Harbour mailing list
Harbour@harbour-project.org
http://lists.harbour-project.org/mailman/listinfo/harbour
