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: [ada] rts support for alpha-linux


> Ada maintainers: I think my versions of sigset_t and other pthreads
> structures is more correct than in the existing s-osinte-linux.ads.

Did you verify the differences in term of size and alignment by e.g.
compiling s-osinte.adb with -gnatR3 ?

> Of primary concern is the alignment of the structures; using char
> for allocating dummy space when the real structure is aligned a-la
> unsigned long is surely a bug.  Of secondary concern, I have no idea
> what "pragma Convention (C, foo)" does when Ada declares an array 
> and the real C type is a structure.  It does seem inconsistent though.

I'll review your changes, although our goal is to minimize the maintenance
of multiple files that are almost identical (such as s-osinte-linux.ads
and s-osinte-linux-alpha.ads), so it is preferrable to find a way to
share the code between all linux ports, as was doneup to now.

Perhaps you could post diffs between the original linux files you started
from and the new alpha variants for analysis ? I am quite sure we can find
a way to keep a single file instead of duplicating.

Same question for a-intnam.ads

> +   system.ads<system-linux-ia64.ads

system.ads is actually an exception to the rule listed above: we do not
want to share system files between targets, this has caused only
confusion in the past trying to do so: we always end up sooner or later
with parameters that are different and that are not properly updated/maintained
because the system file is shared.

Thanks for your interest in building Ada in your build, I'm sure we'll
come up with a new solution in a few rounds.

Arno


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