This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] 25816 Add configure option to disable TLS...



On 2006-01-17, at 18:20, David Daney wrote:


Marcin Dalecki wrote:
On 2006-01-17, at 06:33, David Daney wrote:
+Note that: @option{--enable-tls} does nothing. If configure
+determines that TLS is not supported, this option will not enable it.
This is not good. The user should get what he asks for. In esp. since
last time I checked configure determines it frequently wrong.

Undoubtedly you are correct. My main concern was to fix the regression in the least invasive manner possible while still being able to fully test the patch.


I have no system that supports TLS where configure guesses incorrectly about assembler TLS support. If you have such a system and would like to write and test a patch to support --enable-tls, that would be great.

No problem. There are enough cross build cases in esp. where configure goes wild.
Please forgive me to not enumerate them fully, because I deal with this kind of
problems by circumventing configure in such cases in my build setup
(cat > config.cache <<EOF... basically because autoconf/automake/autogen
isn't reproducible in the real world out there.)
However at least ppc405 with additional TLS support on a 2.4 linux kernel
series hosted on a non linux system is one case where configure goes badly
wrong.


In the mean time I would like to see if my patch can be committed to 4.1 as is, because it fixes a real regression that I think effects quite a few mipsel-linux systems.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]