Re: linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

On 12/27/2010 05:52 PM, Paul Eggert wrote: > I was thinking more along the lines "don't use -xc99=all". Following up on this idea, I installed a change into Autoconf so that Autoconf-generated 'configure' scripts will no longer default to -xc99=all when using Solaris cc. Please see

Re: linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

Paul Eggert wrote: > it makes sense to install that patch regardless of the other > issues we are talking about. OK, I pushed it. Bruno

Re: linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

On 12/27/2010 05:03 AM, Bruno Haible wrote: > Your proposed answer "don't do that" would imply that every library > is distributed in different variants, one for each standards compliance. No, I was thinking more along the lines "don't use -xc99=all". It's clearly a problem to do that, or anything

Re: linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

On 12/27/2010 05:20 AM, Bruno Haible wrote: > OK, to reduce the runtime cost, here's a new proposed patch. It passes the > tests on Solaris 10, with and without "-xc99=all". Yes, thanks, that fixes the performance problem I was worried about, and it makes sense to install that patch regardless of

Re: linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

On Mon, 27 Dec 2010, Bruno Haible wrote: This is not realistic: People are not distributing libraries in this way, and are not even aware for which standard a library was built and tested for. ("file libfoo.so" does not tell. You need "nm libfoo.so | grep values".) Since it seems that evidence

Re: linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

[dropped autoconf and libtool mailing list CCs] Paul Eggert wrote: > the proposed change would appear to place a > significant performance penalty for the (presumably more common) case > of compiling and linking in the default mode.  I would suggest something > like the following patch instead, wi

Re: linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

Paul Eggert wrote: > Given the other problems that ensue on Solaris when one compiles and > links to different standards, the simplest answer may be just "don't > do that". It's not just the __xpg4 and __xpg6 stuff; it's also the > _lib_version stuff: scanf behaves differently depending on which >

Re: linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

[Adding libtool to the CC: list, since Bob indicates there are libtool and autoconf implications as well. The thread starts at .] On 12/26/2010 09:51 AM, Bruno Haible wrote: > So, when libposix becomes reality, it may be compile

linkat, LINK_FOLLOWS_SYMLINKS, and Solaris

Hi Eric, In the linkat implementation (lib/linkat.c), there is the assumption that the value of LINK_FOLLOWS_SYMLINKS depends only on the current machine and platform. But this is not the case: On Solaris 10 systems (which don't have linkat() in libc), the behaviour of the link() function depends