This is the mail archive of the 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]

Re: [PATCH] Fix aliasing miscompilation

>>>>> "Jakub" == Jakub Jelinek <> writes:

    Jakub> If you set it back to -1, it means the restricted pointer
    Jakub> now points to an alias set which is a subset of never used
    Jakub> alias set, so the end result will be that gcc will assume
    Jakub> the restricted pointer cannot ever alias any aggregate of
    Jakub> the type it points to, which is bad.

I'm sorry; I should have said alias set zero.

    Jakub> So, shouldn't alias.c define a new function which would do
    Jakub> exactly what I did in layout_type and frontends would need
    Jakub> to call this when they are really finished with type
    Jakub> layout?

I think we already have a means of doing that.

There was a lot of discussion in the archives.  In the short term, we
really do not want to anything automatic, without the direct
intervention of the front-end.  The reason is that in some front-ends
(again, C++ is this way) there are not necessarily FIELD_DECLs for all
things that can appear in the record type.

Please look through the previous emails (look for an exchange between
Kenner and myself) a while back.


Mark Mitchell         
CodeSourcery, LLC     

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