Removal of c++filt causes g++ testsuite failures

Removal of c++filt causes g++ testsuite failures

[Kaveh R. Ghazi]

Since around 9/20/2002, I've been getting g++ testsuite errors because
c++filt is missing.  E.g.:

http://gcc.gnu.org/ml/gcc-testresults/2002-09/msg00697.html

 > ERROR: g++.dg/abi/rtti1.C: error executing dg-final: couldn't execute "c++filt": no such file or directory
 > UNRESOLVED: g++.dg/abi/rtti1.C: error executing dg-final: couldn't execute "c++filt": no such file or directory
 > ERROR: g++.dg/opt/vtgc1.C: error executing dg-final: couldn't execute "c++filt": no such file or directory
 > UNRESOLVED: g++.dg/opt/vtgc1.C: error executing dg-final: couldn't execute "c++filt": no such file or directory

This continues in 3.3 and 3.4 through the current CVS.

The removal of c++filt was discussed at length starting here:
http://gcc.gnu.org/ml/gcc-patches/2002-09/subjects.html#00271

I see the problem on mips-irix6 and sparc-solaris2, but I guess all
configurations which use native as/ld utils (i.e. not binutils) will
have this issue.  David E also noted it on AIX here:
http://gcc.gnu.org/ml/gcc-patches/2002-09/msg01288.html

The discussion died off without addressing this issue.  I'd like to
settle what to do about this particular problem so that we don't get
bogus noise in the testsuite results.

Would anyone like to suggest something?

		Thanks,
		--Kaveh
--
Kaveh R. Ghazi			ghazi@caip.rutgers.edu

Re: Removal of c++filt causes g++ testsuite failures

[Zack Weinberg]

"Kaveh R. Ghazi" <ghazi@caip.rutgers.edu> writes:

> Since around 9/20/2002, I've been getting g++ testsuite errors because
> c++filt is missing.
...
> The removal of c++filt was discussed at length starting here:
> http://gcc.gnu.org/ml/gcc-patches/2002-09/subjects.html#00271

No conclusion was reached regarding which of the two projects should
provide c++filt.  I consider it incorrect for both to do so.  I have
no opinion on which project should provide it, but I will not change
the status quo unless everyone involved agrees that it should change.
Discussion trailing off isn't agreement.

It seems to me that the test suite should be looking for the mangled
forms of the identifiers; they're not going to change.

zw