Bugs in 7/27 snapshot

[Brad M. Garcia]

On Thu, 6 Aug 1998, Jeffrey A Law wrote:

>   In message < Pine.LNX.4.01.9808060829530.27570-100000@spud.eng.fore.com >you wr ite:
>   > 1) i960 floating point appears to be buggy - I get
>   >    "Internal compiler error" when compiling some code dealing with
>   >    "double".
> I believe Vladimir fixed this problem very recently.  Are you seeing infinite
> recursion when building the test code?

I've never tried building the test programs.  I am building
cross-compilers, and was under the impression that the test suite didn't
work very well for cross-compilers.

But the cases I have seen cause the compiler to crash after a few seconds
of compiling the code (longer than it should have taken to compile the
file).

>   > 2) pragmas for x86 have changed?  Compiling code for i960 has no problems.
>   >    Compiling same code for x86 results in rejecting the pragmas in the
>   >    code.  As an example:
>   >        pragma align 4   <--  complains about unrecognized pragma.
>   >        pragma pack  4   <--  complains about malformed pragma.
>   >    Why is the behavior changed for x86 but not i960?
> pragma behaviors are often target/system dependent.

But earlier versions (up through 1.0.3) have compiled these same
pragmas with no warnings or errors of any kind.  And the egcs snapshot
for an i960-wrs-vxworks target still compiles this code with no warnings
or errors.

Perhaps the problem is this - the i386-wrs-vxworks target is new.  I 
ported 1.0.3 to work for this target myself (by following information
I found on the web).  What should I look for to find out how pragmas
work for a specific target?

> I'm not aware of any changes in behavior for x86 pragmas.


Brad Garcia
   ___/  __ /  __ /  ___/ "Being the Linux of digital media
  __/   /  /  / _/  __/    would be a very good life."
_/    ____/ _/ _| ____/      - Jean-Louis Gassee, CEO of Be, Inc.