This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Possible Fix for PR42644 and PR42130


Hi Sebastian,

Thanks again for the feedback. I made the changes you suggested.

2010/1/29 Sebastian Pop <sebpop@gmail.com>:
> The patch is not ok yet: please remove all the spaces at the end of
> the line.
>
>>> +2010-01-22 ?Ramakrishna Upadrasta <Ramakrishna.Upadrasta@inria.fr>,
>>> + ? ?Tobias Grosser ?<grosser@fim.uni-passau.de>

Done.

> Please align the beginning of the names.

Done

>>> + ? TODO: Get the types using CLooG instead. This enables further
>
> Point, space, space, new sentence.

Done.


>>> + ? ?if (!useless_type_conversion_p(long_long_integer_type_node, type))
>
> After a function name you should put a space before opening a parenthesis.

Done.

>>> + ? ?{
>>> + ? ?/*There is no signed type available, that is large
>
> You need a space between beginning of a comment and the first letter.

OK.

>>> ? ?if (slot && *slot)
>>> - ? ?return ((ivtype_map_elt) *slot)->type;
>>> + ?{
>>> + ? ?type = ((ivtype_map_elt) *slot)->type;
>>> +
>>> + ? ?if (!useless_type_conversion_p(long_long_integer_type_node, type))
>>> + ? ?{
>>> + ? ?/*There is no signed type available, that is large
>>> + ? ? ?enough to hold the original value. ?*/
>>> +
>>> + ? ?gloog_error = true;
>>> + ? ?return long_long_integer_type_node;
>>> + ? ?}
>
> Here you should return the "type", and not fall through and return
> long_long_integer_type_node. ?This could avoid later problems with
> pointer types compared to non pointer long_long_integer_type_node.

OK.

>>> ?}
>
> Have you run the testsuite with at least make -k check
> RUNTESTFLAGS=graphite.exp?

I have, but I am at loss to understand how to properly interpret the
results. (Even without my change there seem to be unexpected
failures.)

>>> + ? ? ?printf ("%d ", ?(int) Ke[j]);
>>> + ? ? ?printf("# ");
>
> You are executing this program, and this would print on stdout: please
> remove the printfs or #ifdef DEBUG them.

OK. I have changed the testcase. I am not sure how would we compare
that miscompile does not happen if there is no printf and comparison
of output. (Maybe I am missing something basic.)

Thanks
Regards
Ramakrishna

Attachment: 0001-Fix-DealII-type-problems.patch
Description: Binary data


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]