sed bugs? (again)

[Paolo Bonzini]

Sorry for not answering earlier...  when I have some time, I'll try to track
these down.  It is not easy to delve into the glibc matcher.  In the
meantime I'm CCing the author and the libc-alpha mailing list

> > doing:
> >   echo 'abcd' | gsed405 's/^\(a*\)*/<&>/' ; echo $?
> > takes some time and returns:
> >   128
> > instead of the logical:
> >   <a>bcd
> >   0
> > returned by ssed360 and my sed.

> >
> > doing:
> >   echo 'abcd' | gsed405 -n 's/E*\(\(\)\2*b*\)/\1!/p'
> > correctly produces
> >   !abcd
> > but
> >   echo 'abcd' | gsed405 -n 's/E*\(\(\)\2*b*\2*\)/\1!/p'
> > prints nothing and returns 128
> >

> > yet another one. doing:
> >   echo 'abbbbc' | gsed405 -n 's/\(b\{2\}\(C*\)\)\1\2/<&>/p'
> > correctly produces:
> >   a<bbbb>cd
> > but:
> >   echo 'abbbbc' | gsed405 -n 's/\(b\{2\}\(C*\)\)\2\1/<&>/p'
> > results in a segmentation fault. The variant:
> >   echo 'abbbbc' | gsed405 -n 's/\(b\(C*\)\)\2\1/<&>/p' ; echo $?
> > fails to match but return 0, and the variant:
> >   echo 'abbbbc' | gsed405 -n 's/\(bb\(C*\)\)\2\1/<&>/p'
> > dumps core as with b\{2\}

Paolo