This is the mail archive of the gcc@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]

Bug in reload when forming inheritable reload information?


Hi,
Recently I found a test program got wrongly reloaded, as reported in PR52804.
As the comment, I think reload_reg_reaches_end_p in reload1.c should
handle case:
The first reload type is RELOAD_FOR_INPADDR_ADDRESS;
the second reload type is RELOAD_FOR_INPADDR_ADDRESS and the reload
register is same as the first one.

In this case, the first reload is corrupted by the second one, but for now
reload_reg_reaches_end_p still returns 1, resulting in forming wrong
inheritable information,
then mis-leading following instructions.

I am working/testing a patch for this right now, also I'd like to get
some confirmation on this issue,
because RELOAD is a well-known complicated pass.

Any comments? Thanks very much.

-- 
Best Regards.


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