Re: [RFA] gdb/testsuite/gdb.c++/ovldbreak.exp

[Fernando Nasser <fnasser at redhat dot com>]

Michael Elizabeth Chastain wrote:
> 
> This patch fixes these problems in gdb.c++/ovldbreak.exp:
> 
>   - Reduces run time from 8405 seconds (!!) to 4 seconds.
>   - Fixes all 41 test failures.
>   - Disambiguates duplicate messages.
> 
> When a user sets a breakpoint on an overloaded function, GDB offers the
> user a submenu to choose the correct function.  The existing ovldbreak.exp
> loses synchronization if one of its tests fails.  It ends up trapped in
> the submenu, issuing fruitless commands and taking dozens of timeouts.
> That's why ovldbreak.exp takes 8405 seconds to execute.
> 
Argh!

> This new ovldbreak.exp breaks out of the submenu whenever necessary,
> so that even when tests fail, it stays synchronized with the GDB prompt.
> It has an execution time of 4 seconds.
> 
Thank you!

> I fixed all 41 test failures.  g++ has changed some of its type encoding,
> such as "unsigned int" versus "unsigned" and "void" versus "".  My test
> accommodates all the forms that the old test accommodates, plus the
> forms that I see with sourceware gcc.
> 
Perfect.  In the future, when the old gcc's become history we could
tighten up
the patterns again.  But I don't think it matters, because the idea of
the test
is not get the exact output format but rather if the right options were
given
to the user.

> I changed the test messages to emit the same messages for PASS and FAIL,

You got the spirit!  Thanks!

> and I disambiguated some duplicate test messages.
> 

Yes, yes yes!


> I modularized the file with more procedures, and I documented the procedures.
> I removed excessive calls to "info break".  The old file was 583 lines long;
> the new file is 349 lines long.
> 
Nice.


> I tested this on:
> 
>   Red Hat Linux 6.2, native, with native gcc (egcs-2.91.66)
>   Red Hat Linux 6.2, native, with sourceware cvs gcc
>   Red Hat Linux 7, native, with native gcc (gcc 2.96)
>   Red Hat Linux 7, native, with sourceware cvs gcc
> 
> OK to commit?
> 

Yes, please.

And thank you for doing this.


P.S: It was quite hard to review your patch due to the size, patterns
etc.
As far as I can tell it looks just perfect.  Also, this should not be
platform
dependent so I guess your tests are enough (despite being just Linux
native).
IMO, in these cases there is nothing much we can do but to check it in
and keep 
a close watch on it. 


-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9