Re: [Mingw-msys] Regression in MSYS

[Earnie B. <ear...@ya...>]

Paul G. wrote:

>Umm, I think that there was a change in the make.exe being used by Msys prior to 1.0.8.  As I recall, and I hope 
>someone will be so kind as to correct me on this if I am wrong, the older version of make.exe was not working in 
>most cases, thus justifying the removal of the old make.exe.
>
>	Paul G.
>  
>
Has nothing at all to do with the problem.  All versions of MSYS are 
using the same version of make.

>On 30 Aug 2003 at 18:32, Daniel Herring wrote:
>
>  
>
>>Hi all,
>>
>>I'm trying to compile Coin3D (latest cvs, www.coin3d.org) using MinGW
>>2.0.0-3 and MSYS on a WinXP platform.
>>
>>In short, the build fails on recent versions of MSYS, but succeeds on
>>older ones.
>>
>>Failed:
>>MSYS-1.0.10-rc-1.exe
>>MSYS-1.0.9.exe
>>MSYS-1.0.8.exe
>>
>>Success:
>>MSYS-1.0.6-i386.exe
>>
>>Mixed:
>>MSYS-1.0.7-i386.exe
>>(This version did not fail where the others did; I must have done
>>./configure wrong or something.)
>>
>>All these builds are from http://prdownloads.sf.net/mingw/
>>
>>To recreate,
>>$ cvs -d:pserver:cv...@cv...:/export/cvsroot login
>>(password: cvs)
>>$ cvs -d:pserver:cv...@cv...:/export/cvsroot co Coin-2
>>$ cd Coin-2
>>$ ./configure --path=C:/Coin3D
>>$ make
>>
>>Notes:
>>1.) Coin-2/src/shapenodes/SoText2.cpp seems to have an error; I change
>>line 686 to
>>  this->stringwidth.append((int)w);
>>
>>2.) I actually did the cvs checkout using the client on
>>www.TortoiseCVS.org
>>
>>3.) Unfortunately, this build takes a long time to complete.
>>
>>
>>
>>***** MSYS 1.0.9 *****
>>
>>"make" crashes with
>>
>>rm -fr .libs/libCoin.lax/libfields.a
>>mkdir .libs/libCoin.lax/libfields.a
>>(cd .libs/libCoin.lax/libfields.a && ar x
>>/c/CoinSRC/Coin-2/src/fields/.libs/libfields.a)
>>      0 [main] sh 3224 open_stackdumpfile: Dumping stack trace to
>>sh.exe.stackdump ../libtool: line 1:  3224 Segmentation fault     
>>(core dumped) ar x $xabs make[3]: *** [libCoin.la] Error 139 make[3]:
>>Leaving directory `/c/CoinSRC/Coin-2/src' make[2]: *** [all-recursive]
>>Error 1 make[2]: Leaving directory `/c/CoinSRC/Coin-2/src' make[1]:
>>*** [all-recursive] Error 1 make[1]: Leaving directory
>>`/c/CoinSRC/Coin-2' make: *** [all] Error 2
>>
>>
>>Here's the stack trace (if anyone can decipher it):
>>MSYS-1.0.9 Build:2003-07-03 07:26
>>Exception: STATUS_ACCESS_VIOLATION at eip=71098795
>>eax=0A0E5288 ebx=710A5FB4 ecx=00002CE0 edx=016FC48D esi=0A0E25A8
>>edi=710A5CA4
>>ebp=0022A46C esp=0022A434 program=C:\msys\bin\sh.exe
>>cs=001B ds=0023 es=0023 fs=0038 gs=0000 ss=0023
>>Stack trace:
>>Frame     Function  Args
>>0022A46C  71098795  (710A5020, 00002CDC, 77E77F57, FFFFFFFF)
>>0022A49C  71031350  (00002CDC, 00000000, 00000000, 00002BD8)
>>0022A4BC  710310D8  (00002CDC, 0A0CB6D0, 00002BD9, 00000000)
>>0022D44C  7104375B  (0A0CB6D0, 0A0C0B48, 00000001, 71061DFE)
>>0022E80C  71061EED  (00000000, 0A018158, 0A0153B0, 0A0BB638)
>>0022E84C  710639FA  (00000000, 00000003, 0A018158, 0A0153B0)
>>0022E87C  710114DB  (0A018158, 0A0153B0, 0A0BB638, 710A5020)
>>0022E89C  7108D872  (0A018158, 0A0153B0, 0A0BB638, 0022E908)
>>0022E96C  004127A4  (0A018158, 0A0153B0, 0A0BB638, 0042B3C4)
>>0022E9C8  00412706  (0A0BBB08, 00000000, 0A0B94B0, FFFFFFFF)
>>0022EA48  00411CA7  (0A0BE190, FFFFFFFF, FFFFFFFF, 00000000)
>>0022EAA8  0040E9C6  (0A0BE178, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EAF8  0040FD6A  (0A0BAB58, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EB58  0040EBE2  (0A0BAB58, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EBA8  0040F693  (0A0BAB80, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EC14  0040E68F  (0A0BAB80, 00000000, FFFFFFFF, FFFFFFFF)
>>End of stack trace (more stack frames may be present)
>>
>>
>>The annoying part is that the MSYS shell does not terminate and I can
>>execute this "ar x" command manually without any errors.
>>    
>>
This has to do with many invocations of the shell through make causing a 
bug I've yet to find to be raised in the malloc routines for the 
environment.  Version 1.1.0 will have this resolved (the code raising 
the bug has been rewritten, the bug still exists).  As soon as I can 
catch my breath from the email virus that is in effect and the heavier 
load from the real work environment I'll put up a snapshot.

>>
>>***** MSYS 1.0.8 *****
>>
>>This version dies later...
>>
>>rm -fr .libs/libCoin.lax/libfields.a
>>mkdir .libs/libCoin.lax/libfields.a
>>(cd .libs/libCoin.lax/libfields.a && ar x
>>/c/CoinSRC/Coin-2/src/fields/.libs/libfields.a)
>>rm -fr .libs/libCoin.lax/libglue.a
>>mkdir .libs/libCoin.lax/libglue.a
>>(cd .libs/libCoin.lax/libglue.a && ar x
>>/c/CoinSRC/Coin-2/src/glue/.libs/libglue.a)
>>rm -fr .libs/libCoin.lax/libmisc.a
>>mkdir .libs/libCoin.lax/libmisc.a
>>(cd .libs/libCoin.lax/libmisc.a && ar x
>>/c/CoinSRC/Coin-2/src/misc/.libs/libmisc.a)
>>rm -fr .libs/libCoin.lax/liblists.a
>>mkdir .libs/libCoin.lax/liblists.a
>>(cd .libs/libCoin.lax/liblists.a && ar x
>>/c/CoinSRC/Coin-2/src/lists/.libs/liblists.a)
>>rm -fr .libs/libCoin.lax/libnodekits.a
>>mkdir .libs/libCoin.lax/libnodekits.a
>>(cd .libs/libCoin.lax/libnodekits.a && ar x
>>/c/CoinSRC/Coin-2/src/nodekits/.libs
>>/libnodekits.a)
>>rm -fr .libs/libCoin.lax/libnodes.a
>>mkdir .libs/libCoin.lax/libnodes.a
>>(cd .libs/libCoin.lax/libnodes.a && ar x
>>/c/CoinSRC/Coin-2/src/nodes/.libs/libnodes.a)
>>../libtool: /mingw/bin/ar: Invalid argument
>>make[3]: *** [libCoin.la] Error 126
>>make[3]: Leaving directory `/c/CoinSRC/Coin-2/src'
>>make[2]: *** [all-recursive] Error 1
>>make[2]: Leaving directory `/c/CoinSRC/Coin-2/src'
>>make[1]: *** [all-recursive] Error 1
>>make[1]: Leaving directory `/c/CoinSRC/Coin-2'
>>make: *** [all] Error 2
>>
>>
>>Again, MSYS does not crash and I can manually do
>>$ cd src/.libs/libCoin.lax/libnodes.a/
>>$ ar x /c/CoinSRC/Coin-2/src/nodes/.libs/libnodes.a
>>without error.
>>    
>>
I'm not sure where this problem lies.  What version of libtool is used 
by Coin?

>>
>>***** MSYS 1.0.10-rc1 *****
>>
>>This version gets a little further before crashing.
>>
>>rm -fr .libs/libCoin.lax/libnodes.a
>>mkdir .libs/libCoin.lax/libnodes.a
>>(cd .libs/libCoin.lax/libnodes.a && ar x
>>/c/CoinSRC/Coin-2/src/nodes/.libs/libnodes.a)
>>rm -fr .libs/libCoin.lax/libshapenodes.a
>>mkdir .libs/libCoin.lax/libshapenodes.a
>>(cd .libs/libCoin.lax/libshapenodes.a && ar x
>>/c/CoinSRC/Coin-2/src/shapenodes/.libs/libshapenodes.a)
>>rm -fr .libs/libCoin.lax/libprojectors.a
>>mkdir .libs/libCoin.lax/libprojectors.a
>>(cd .libs/libCoin.lax/libprojectors.a && ar x
>>/c/CoinSRC/Coin-2/src/projectors/.libs/libprojectors.a)
>>      0 [main] sh 808 open_stackdumpfile: Dumping stack trace to
>>sh.exe.stackdump
>>../libtool: line 1:   808 Segmentation fault      (core dumped) ar x
>>$xabs make[3]: *** [libCoin.la] Error 139 make[3]: Leaving directory
>>`/c/CoinSRC/Coin-2/src'
>>
>>Here's the stack trace
>>MSYS-1.0.10 Build:2003-07-29 09:21
>>Exception: STATUS_ACCESS_VIOLATION at eip=71097DBB
>>eax=012E0120 ebx=0A0CC568 ecx=00002008 edx=0A0E4910 esi=0A0CC560
>>edi=0A0CE568
>>ebp=0022D3FC esp=0022D3C4 program=C:\msys\bin\sh.exe
>>cs=001B ds=0023 es=0023 fs=0038 gs=0000 ss=0023
>>Stack trace:
>>Frame     Function  Args
>>0022D3FC  71097DBB  (710A5020, 0A0CC568, 00003000, 71031309)
>>0022D42C  710313EF  (0A0CC568, 00003000, 0022D3EC, 0A0D4511)
>>0022D44C  71031138  (0A0CC568, 00003000, 00004CF2, 71061DFE)
>>0022E80C  71061F6B  (00000000, 0A0B9440, 0A0BF788, 0A0BB638)
>>0022E84C  710639FA  (00000000, 00000003, 0A0B9440, 0A0BF788)
>>0022E87C  710114DB  (0A0B9440, 0A0BF788, 0A0BB638, 710A5020)
>>0022E89C  7108D872  (0A0B9440, 0A0BF788, 0A0BB638, 0022E908)
>>0022E96C  004127A4  (0A0B9440, 0A0BF788, 0A0BB638, 0042B3C4)
>>0022E9C8  00412706  (0A018618, 00000000, 0A010F28, FFFFFFFF)
>>0022EA48  00411CA7  (0A012FF8, FFFFFFFF, FFFFFFFF, 00000000)
>>0022EAA8  0040E9C6  (0A012FE0, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EAF8  0040FD6A  (0A0BE128, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EB58  0040EBE2  (0A0BE128, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EBA8  0040F693  (0A0BE140, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EC14  0040E68F  (0A0BE140, 00000000, FFFFFFFF, FFFFFFFF)
>>0022EC64  004499BB  (0A0BB180, 0044C464, 00000004, 0042BFC7)
>>End of stack trace (more stack frames may be present)
>>
>>    
>>
An attempt to work around the 1.0.9 problem.  It worked around the 
problem in some instances but not all.

>>Thus, it appears that something broke between version 1.0.7 and 1.0.8
>>I'm willing to provide more information if you can tell me what you
>>need.
>>    
>>

I don't know what else you can provide.  I hope to have a snapshot up of 
1.1.0 soon.

Earnie.