Re: More Objective C @"string" bugs...

[Geoff Keating <geoffk at geoffk dot org>]

Neil Booth <neil@daikokuya.demon.co.uk> writes:

> Joseph S. Myers wrote:-
> 
> > GCC developers follow the standard procedure of running the testsuite and
> > avoiding regressions - it's simply that the ObjC testsuite is very small
> > so problems are likely not to be shown up.  Perhaps there are Objective C
> > testcases in the old archives of bug-gcc, egcs-bugs and gcc-bugs that
> > could be put into the testsuite, so much as to improve its coverage as to
> > test for regressions on old bugs?
> 
> Well, the Objective C testsuite went from 2 regressions last summer, to 3,
> to 5, to all-but-a-couple, and no-one seemed to care.  I wondered why
> the committers of whatever patches caused it didn't notice.

If the problems had been generic, the automated tester would have
noticed.  I believe that the main cause of new objc test failures is
new objc testcases.

> I hope a lot of these problems can be avoided by more clearly defining the
> interface between different phases of the compiler.  This appears to be
> the direction most people want to go in, and is slowly happening, but
> progress is (naturally) slow.

objc is a little bit more special, because of its close relationship
with the C frontend.

-- 
- Geoffrey Keating <geoffk@geoffk.org>