Re[2]: [Gc] 7.3alpha3 not finishing test on MingW with c++ options

Ivan Maidanski ivmai at mail.ru
Tue Oct 30 12:39:15 PST 2012


Hi Joshua,

Thank you.

I've managed to reproduce test_cpp crash (I haven't tried with the other test though):
C:\MinGW\bin\gcc -I include -I libatomic_ops/src -g -DGC_THREADS -c extra/gc.c
C:\MinGW\bin\g++ -I include -I libatomic_ops/src -g -DGC_THREADS tests/test_cpp.cc gc.o -static-libgcc
a.exe 100

The crash is at some random iteration but always at "delete a" (where 'a' is of uncollectable type). Unfortunately, gdb does not show backtrace at crash point. Should be investigated more.

realloc_test does not contain any forking, it only call GC_malloc many times.

Regards,
Ivan

Mon, 29 Oct 2012 21:36:20 +0000 (UTC) Joshua Scholar <JoshuaScholar at gmail.com>:
>	
>
>
	
	
>
		
		
			
>Since you couldn't reproduce I decided to make sure that _I_ can 
>
reproduce.
>
>
I'm getting hangs at random (a few times in gctest and once in 
>
realloc_test), 
>
and test_cpp ALWAYS CRASHES with a windows exception that reads:
>
Problem signature:
>
  Problem Event Name:	APPCRASH
>
  Application Name:	test_cpp.exe
>
  Application Version:	0.0.0.0
>
  Application Timestamp:	508ecb56
>
  Fault Module Name:	ntdll.dll
>
  Fault Module Version:	6.1.7601.17725
>
  Fault Module Timestamp:	4ec49b60
>
  Exception Code:	c0000005
>
  Exception Offset:	00052c39
>
  OS Version:	6.1.7601.2.1.0.256.1
>
  Locale ID:	1033
>
  Additional Information 1:	0a9e
>
  Additional Information 2:	0a9e372d3b4ad19135b953a78882e789
>
  Additional Information 3:	0a9e
>
  Additional Information 4:	0a9e372d3b4ad19135b953a78882e789
>
>
>
gctest is currently hung, though it passed 4 times before it hung.
>
I ran in manually after that.  It ran twice then hung on the third 
>
try.
>
>
I put the laptop in suspend in the middle of reallloc_test once by 
>
accident.  When i unsuspended there were two realoc_test processes 
>
and they were both hung.  While realloc_test doesn't seem to hang 
>
tested on it's own, I did try starting a bunch of them at once 
>
manually, ie "realloc_test&" that worked once.  But on the second 
>
try I was left with two processes that don't end. Does realloc_test 
>
fork? Whenever I run it I see two processes - but I kill one and they both 
>
disappear.
>
>
I'm running on an atom with hyperthreading so if the tests use multiple 
>
threads, then I am getting something like multiprocessor contention.
>
>
My MinGW compiler is the 4.5.4 which is the current default install. 
>
Does the linux environment I ran autogen.sh on affect things?  I think 
>
that one had gcc 4.2
>
>
I DID try your incantation by the way the resulting file runs, but
>
as I said I haven't been able to get middle to crash anyway. The 
>
first time I ran your versionthe final heap size was 65536, the 
>
other 3 times I ran it the final heap size was 131072.
>
>
since I didn't figure out how gctest is compiled, I tried your 
>
incantation with test_cpp.cc instead of middle.c, but that failed 
>
to compile with errors like:
>
C:\...\AppData\Local\Temp\ccE0Mpa1.o:test_cpp.cc:(.text+0x2fc): 
>
undefined reference to `operator delete(void*)'
>
C:\...\AppData\Local\Temp\ccE0Mpa1.o:test_cpp.cc:(.rdata$_ZTI1C
>
[typeinfofor C]+0x0): undefined reference to `vtable for 
>
__cxxabiv1::__vmi_class_type_inf
>
o'
>
>
>
_______________________________________________
>
Gc mailing list
>Gc at linux.hpl.hp.com
>http://www.hpl.hp.com/hosted/linux/mail-archives/gc/
>
			
		
		
	

	
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://napali.hpl.hp.com/pipermail/gc/attachments/20121031/4bef8159/attachment.htm


More information about the Gc mailing list