[Bug c++/11489] [3.4 Regression] ICE/segfault on legal code

[numerical dot simulation at web dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11489



------- Additional Comments From numerical dot simulation at web dot de  2003-07-10 15:12 -------
Subject: Re:  [3.4 Regression] ICE/segfault on legal code

bangerth at dealii dot org wrote:

> Regarding the state of gcc: First, you are testing mainline, not a released
> version.

... because there is no gcc release that is able to compile my code :-(
Here I am again in front of an idle  $30000 PA-RISC machine waiting for
a compiler to appear.

> We just closed the development phase, so some instability needs
> to be expected at this point in time.

OK.

> Second, about what Andrew said: it's
> not quite true that gcc gas no decent testsuite. For one, the testsuite has
> >20,000 testcases. Secondly, several people build their (large) C++
> packages every night with present CVS versions of gcc. But if you asked
> developers to do that for themselves before check-in,

No, I only ask the C++ parser guys to run a decent subset,
namely some 50000 C++ lines, which cannot take too long.

> it would take hours
> or days (my nightly builds, for example, take 2-3 hours).

IMHO a 3 hour test before checkin is worth the cost:
I lost several hours today just to check things the breaker
of the code could have catched and understood immediately.

If I'd fiddle around in the template parser stuff I'd check that
on some code before checkin. It is a matter of discipline, not
of possibilities. Why do _I_ have to waste a whole day on this bug?

Or do you wish me to wait until release?


> There needs to be
> a balance between the effort to develop gcc and our willingness to accept
> that once in a while something gets broken.

Agreed. I just cannot see a balance right now.
Alas, not once-in-a-while, but every-day-something-else gets broken.


Markus