This is the mail archive of the binutils@sourceware.org mailing list for the binutils 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: binutils version of snapshot builds


On Wed, 21 Mar 2007, Matthias Klose wrote:

> That doesn't change the snapshot builds, I still get "070315.20070315".

Then I think there's some bug in the snapshot generation process, 
"2.17.50.20070315" would be appropriate not "070315.20070315".

> Configuring a snapshot build --with-pkgversion="Ubuntu 2.17.20070321cvs-1" (the
> latter beeing the package version) I end up with a version string:
> 
>   "GNU ld (Ubuntu 2.17.20070321cvs-1) 2.17.50.20070321"
> 
> That doesn't look robust with the current practice of scanning the ld version.

It conforms to the GNU Coding Standards definition of version numbers.  
GCC's ld version number checks have been fixed (libstdc++ had hardcoded 
checks for the word "version" that used to be there).

> Would it be more appropriate to append the package version information like it
> is done in gcc? For now I avoid adding the package version into the PKGVERSION
> string.

The current GCC practice, "4.3.0 20070319 (experimental)" at the end, does 
not conform to the GNU Coding Standards.

Appending to the end of the string, as well as not conforming to the GNU 
Coding Standards, breaks other version checks.  The checks in glibc for as 
and ld versions, for example, extract the last version number found.  
Note that if PKGVERSION includes a version number that number may relate 
to a larger package and have nothing to do with GNU binutils version 
numbers, so appending to the end is liable to break the glibc version 
checks.

> Unrelated to your patch: Building the snapshot using an earlier snapshot having
> the same sonames for the shared libraries fails with an as segfault for the
> first file in opcodes. This could be avoided including the date string in the
> soversion (bfd/configure.in). Are there reasons against this configuration
> (besides ending up with a version string "2.17.50.20070321.20070321")?

I think having 2.17.50.20070321 in the SONAME is reasonable.

-- 
Joseph S. Myers
joseph@codesourcery.com


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