Re: pkg/43239: lang/python26 dumps core during "make test" after upgrade to 2.6.5

["David H. Gutteridge" <dhgutteridge%sympatico.ca@localhost>]

The following reply was made to PR pkg/43239; it has been noted by GNATS.

From: "David H. Gutteridge" <dhgutteridge%sympatico.ca@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: pkg/43239: lang/python26 dumps core during "make test" after 
upgrade to 2.6.5
Date: Mon, 3 May 2010 22:40:27 -0400

 On 2010-05-03, at 12:55 AM, David Holland wrote:
 >>>Clearly it shouldn't dump core. But, isn't pyexpat built as a separate
 >>>package?
 >>
 >>Yes, there's a separate package for pyexpat, but apparently it gets
 >>built as part of the standard Python modules, so it's included in the
 >>default test coverage.
 >
 >That may be a bug... the impression I have is that it's supposed to be
 >disabled.
 
 When the package was modified to exclude pyexpat from the build, the
 test suite coverage wasn't correspondingly altered, so when the test
 runs, it tries to access a module that doesn't exist.  (It turns out
 that issue has nothing to do with the upgrade to 2.6.5, nor to the
 other expat issues surrounding changes to mk/buildlink3/find-files.mk.)
 
 I'm not sure if this is the "pkgsrc way" to do this, but my patch
 below fixes the test coverage mismatch.  (Alternately we could patch
 the Python make file.  My way was faster to test.)
 
 --- Makefile.old        2010-05-02 14:28:54.000000000 -0400
 +++ Makefile    2010-05-03 20:21:26.000000000 -0400
 @@ -144,6 +145,7 @@
  PLIST_SUBST+=  PY_VER_SUFFIX=${PY_VER_SUFFIX:Q}
  
  TEST_TARGET=   test
 +MAKEFLAGS+=    EXTRATESTOPTS="-x test_pyexpat"
  INSTALL_TARGET=        altinstall
  
  REPLACE_INTERPRETER+=  py26
 
 >>I'm using pyexpat so I haven't risked installing 2.6.5 and
 >>rebuilding pyexpat to see if the separate package is similarly
 >>affected.  (But I can try testing this tomorrow in a VM image.)
 >
 >I can test it (though on amd64) if you tell me what to do...
 
 My test that the pyexpat (or py-expat, to pkgsrc) module works was
 simply the part of the coverage suite that was failing
 (Lib/test/test_pyexpat.py).  (Well, that and some code of mine that
 uses pyexpat, but the test that Python provides is far more thorough.)
 I got optimistic, went ahead and upgraded to 2.6.5 and rebuilt
 py26-expat as well, and test_pyexpat.py is causing the interpreter to
 seg fault.  I think this is a separate issue...  I'm continuing to
 test at my end.
  
 Dave