This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH][Ada] XFAIL ACATS c380004 and c953002, update and import tests from ACATS 2.5L
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- To: roger at eyesopen dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 10 Jan 05 14:52:28 EST
- Subject: Re: [PATCH][Ada] XFAIL ACATS c380004 and c953002, update and import tests from ACATS 2.5L
I believe you're mistaken! The size of the type created by Gigi
doesn't overflow, but GNAT's poor estimates of the bounds on the
type's size, as given by min_size and max_size do overflow. Previous
versions of GCC were able to provide more accurate size bounds on the
type (which is why c38004 passed with earlier versions of GCC), but
this analysis depended upon fold to perform transformations that were
often not beneficial to code generation.
Nope. The type produced by the front end is definitely wrong and has
always been. All versions of GNAT since 1.0 have handled this
"incorrectly" (as defined by this ACATS test). It would work by
accident on some 32-bit systems one in a while if the overflow was
missed, but the type is most definitely wrong. The RTS/Front end
folks can tell you more here.
Arno: can you explain the issue here?