On Wed, 10 Mar 2010, Szak�ts Viktor wrote:

Hi,

> So I assume it's safe to add it for gcc on other platforms 
> as well. Any limitations here? F.e. does OS/2 or DJGPP support it?

I have not idea how it works in non *nix builds.
In DJGPP pipes are emulated by normal files and DOS is single process
environment so using this switch here can only seriously reduce the
performance. The version of GCC I have does not support it.
BTW now when we added -j<n> support in GNU make using -pipe does not
reduce the Harbour build time. Just simply we effectively use all
CPU power setting <n> to value greater then number of CPUs but in
the past the difference were noticeable.

> > Maybe you only should check it for hbmk2 and if necessary add code
> > to explicitly delete unfinished destination files on errors.
> I did a very simple test with one good and one broken .c input 
> file (with mingw only), but couldn't spot a difference in behavior 
> when using -pipe. The .o file for the right one was created, 
> for the wrong one, it wasn't.

AFAIR it was always working without any problems in simple tests
when small files were used. The problems appeared only for long files
when backends begin to process incoming data and write the results
and the compilation was interrupted due to errors.
It's possible that it is fixed in current GNU tools anyhow it maybe
be hard to well test it.

best regards,
Przemek
_______________________________________________
Harbour mailing list (attachment size limit: 40KB)
[email protected]
http://lists.harbour-project.org/mailman/listinfo/harbour

Reply via email to