On Fri, Nov 2, 2012 at 5:28 PM, Branko Čibej <br...@wandisco.com> wrote: > On 02.11.2012 14:21, Ivan Zhakov wrote: >> On Fri, Nov 2, 2012 at 5:12 PM, Branko Čibej <br...@wandisco.com> wrote: >>> On 02.11.2012 12:36, Ivan Zhakov wrote: >>>> On Wed, Oct 31, 2012 at 2:04 AM, Branko Čibej <br...@wandisco.com> wrote: >>>> [..] >>>>> The sysinfo bits have static (build-time) info and dynamic (runtime) >>>>> info. Presumably the only difference will be noticing when you're >>>>> running the program on a different "size" of OS, e.g., running 32-bit >>>>> code on a 64-bit OS (hopefully in some compatibility mode). >>>>> >>>>> For the purpose of user agent strings, the host triplet exposed in the >>>>> #define in svn_private_config.h should be more than good enough. >>>>> >>>> I agree that using autoconf to collect OS type is much better. But I >>>> think we should use $target, instead of $host for user-agent to >>>> support cross compile scenarios. Also I've checked $target_os for our >>>> build bots and their values are: >>>> * OpenBSD: 'i386-unknown-openbsd5.0' ($target_os = 'openbsd5.0', >>>> $target_vendor='unknown') >>>> * Centos: 'x86_64-redhat-linux-gnu' ($target_os='linux-gnu'; >>>> $target_vendor='redhat') >>>> * Ubuntu: ''x86_64-unknown-linux-gnu'' ($target_os=''linux-gnu''; >>>> $target_vendor='unknown') >>>> >>>> I've also googled for different autoconf outputs: >>>> * MacOS: 'x86_64-apple-darwin11.2.0' >>>> * cygwin: 'i686-pc-cygwin' >>>> * mingw: 'i686-pc-mingw32' >>>> >>>> We can use all $target triplet in user-agent or just $target_os. I >>>> have no opinion on this matter. Any thoughts? >>> I considered that, but our build scripts very definitely do not support >>> cross-compiling. So anyone who tries that and succeeds can easily add >>> another line to the already-huge patch that made cross-compiling possible. >>> >> Using $target or $host is not important question for me. More >> important question which components use in user-agent string: all >> triplet or only OS? > > I suggest using the same string that's displayed by svn --version, and > that's the whole triplet. > I'm fine with this approach. I'll wait for other opinions on this matter and finish my patch. Thanks!
-- Ivan Zhakov
