[libgfortran] patch PR18857

Paul Brook paul@codesourcery.com
Fri Apr 29 00:24:00 GMT 2005


On Thursday 28 April 2005 13:11, THOMAS Paul Richard 169137 wrote:
> Is the testsuite program too much?  It could be reduced to two cases, in
> order to test the disappearance of the source of the PR.

I would add tests for:
- A whole array (eg. a = matmul(b, c))
- an array section (eg. a = matmul(b(1::2, :), b(:, 1:N/2))
- An argument that requires a temporary (eg. a = matmul (a, b+c))
- Returning to a temporary (eg. a = b + matmul(c, d))

> 2005-04-28  Paul Thomas  <pault@gcc.gnu.org>
>
> 	PR libfortran/18857
> 	* generated/matmul_c4.c:
> 	* generated/matmul_c8.c:
> 	* generated/matmul_i4.c:
> 	* generated/matmul_i8.c:
> 	* generated/matmul_r4.c:
> 	* generated/matmul_r8.c:
> 	* m4/matmul.m4: Removes assertions that component base of the
> 	input and output gfc_array structures should be 0 for all 7
> functions.

Ok.

Changelog should read something like:

 	* m4/matmul.m4: Remove incorrect assertions.
 	* generated/matmul_c4.c: Regenerate
 	* generated/matmul_c8.c: Regenerate
 	* generated/matmul_i4.c: Regenerate
 	* generated/matmul_i8.c: Regenerate
 	* generated/matmul_r4.c: Regenerate
 	* generated/matmul_r8.c: Regenerate

Don't forget to add the ChangeLog entry for the testcase.

There's no need to include generated files in the diff you send to the list, 
They just make it harder to read.

Paul



More information about the Gcc-patches mailing list