gccbug reports?

[Steve Kargl]

On Mon, Jun 14, 2004 at 10:58:31PM -0400, Daniel Berlin wrote:
> 
> On Jun 14, 2004, at 10:43 PM, Steve Kargl wrote:
> 
> >On Tue, Jun 15, 2004 at 04:40:24AM +0200, Giovanni Bajo wrote:
> >>Steve Kargl wrote:
> >>
> >>>The "Category" for 2 of the missing reports was "fortran" and the
> >>>3rd report's "Category" was "optimization".
> >>
> >>I forgot to update gccbug when splitting optimization into
> >>rtl-optimization/tree-optimization in Bugzilla. Fixed thus, committed 
> >>to trunk,
> >>3.4 and 3.3 as rubber-stamped by Zack.
> >>
> >Please see the PR that started this thread.  From the info
> >there, tell me how to choose between rtl- and tree-?
> 
> How do they know it's an optimization problem in the first place?

They don't

> Just because your case works at -O1 and fails at -O2 (or works at -O0 
> and fails at -O1), doesn't mean the problem lies in one of the 
> optimizers.

But, this is a pretty good indication that requesting a higher
optimization level is causing a problem and the problem *may be*
in an optimizer.

> How would they know it's a "middle-end" problem, for example?
> The categories are *not* for users to really know about.

Then, the categories/components should not be exposed to the normal
user in gccbug or on bugzilla.


> The web interface also has a description of each category if you click 
> on the word "components" next to the drop down list of components.

RSI.

> There are few users who use the email interface (though they seem to 
> consistently do so), which has no descriptions of the components there.

I use the email interface because I can attach patches to a bug report.
If you go to bugzilla and create a new bug report, there is no way
to *attach* (i.e., preserve tabs) a patch.

-- 
Steve