Patch to fix irix5 & irix6 libstdc++-v3 scan_not() implementations

Kaveh R. Ghazi ghazi@caip.rutgers.edu
Mon May 13 07:20:00 GMT 2002


 > From: Benjamin Kosnik <bkoz@redhat.com>
 > 
 > > Okay to install on the trunk?
 > 
 > Yes, with changes.
 > 
 > > 2002-05-10  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
 > > 
 > > 	* config/os/irix/irix5.2/bits/ctype_inline.h (scan_is,
 > > 	scan_not): Fix typo, use this->is() rather than manually (and
 > > 	perhaps incorrectly) inlining it.
 > > 	* config/os/irix/irix6.5/bits/ctype_inline.h (scan_is,
 > > 	scan_not): Likewise.
 > 
 > OK.
 > 
 > > 	* testsuite/22_locale/ctype_scan.cc: New file.
 > 
 > This is ok, but you'll need to name it ctype_scan_char.cc.

Thanks, done.

 > There's an existing file, ctype_members_char that would include
 > this functionality (which sadly has no functional test for
 > scan... whoops), but your tests are pretty extensive so we might as
 > well break it out now, since you've been kind enough to code up
 > nice tests.

I did see ctype_members_char and considered adding my test to that
file, however I noted that it is currently contains this:

	// XXX This test (test02) is not working for non-glibc locale models.
	// { dg-do run { xfail *-*-* } }

This makes adding my test to it useless as a mechanism for catching
broken scan implementations.

IMHO, the convention of adding test#() functions to existing
libstdc++-v3 tests is really a bad idea, since you lose every test in
the file when it's marked xfail like this.  We should consider
breaking out every test#() function into its own main() in its own
file.  Or at least do so for tests marked xfail so the other test
functions in that file get checked properly and appear as real
failures if they break.


 > Usually, a wide test is added too (see the rest of the tests in that
 > directory.) Since wide support is off on Irix, it might make sense for
 > me to do that after you've committed this. If you'd like, I can add that
 > later, or you can do it now, either way it's ok.

Yes, I'll have to pass on the wchar test.  So I accept your offer to
do it instead. :-)

		Thanks,
		--Kaveh
--
Kaveh R. Ghazi			Director of Systems Architecture
ghazi@caip.rutgers.edu		Qwest Global Services



More information about the Gcc-patches mailing list