Am Sat, 17 Jul 2010 11:04:31 +0900 schrieb Takashi Ono <[email protected]>:
> I think we should decide whether or not to give up precompiled header > support on gcc. That decision should be postponed to the new GNU Make build system. It makes no sense to fix gcc pch in the old system. In the new GNU Make system, there is currently no pch support on gcc although it can be added. > Precompiled header support is marked as EXPERIMENTAL in configure > script for a very long time. On Windows, the buildbot is using pch > and the featue is well tested these days. Windows pch and gcc pch are really very different beasts. > Pch on gcc requires that the headers can be compiled in both > exception handling enabled and disabled and that the precompiled > header is included not more than once in each compilation unit. Some > new code violates these conditions from time to time and I am raising > issues like i#11379. The issue looks unrelated to me. However, pch needs to have one file with exceptions and one without on Windows too, so no difference there. > As the conditions for writing code with precompiled headers are not > well documented and pch on gcc is not so stable, these issues take > long time to be fixed. That might be true. > I feel like I am the only one using pch on gcc. > If we do not want to keep pch feature on gcc, we should remove it in > configure. IMHO the question is more like: Is it worth it to add the feature to the new build system? Is there any noticable performance improvement by pch on gcc platform or does it not matter anyways as these have decent filecaching? Is pch halfway stable in recent gcc releases? BR, Bjoern --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
