'seen' branch identifies as v2.48.0

Hi Junio,

When building git from tonight's seen branch, it claims to be version 2.48.0, thus:

  $ pwd
  /home/ramsay/git
  $ 
  
  $ ./git version
  git version 2.48.0
  $ 
  
  $ git describe
  v2.48.0-246-g568b3e75e9
  $ 
  
  $ git log -1
  commit 568b3e75e9cc62870cace4e290d7c2bd3241e3e5 (HEAD -> seen, origin/seen)
  Merge: 1a6fedb381 20d77dd0a5
  Author: Junio C Hamano <gitster@pobox.com>
  Date:   Mon Jan 13 10:20:35 2025 -0800
  
      Merge branch 'ps/zlib-ng' into seen
      
      * ps/zlib-ng:
        compat/zlib: allow use of zlib-ng as backend
        git-zlib: cast away potential constness of `next_in` pointer
        compat/zlib: provide stubs for `deflateSetHeader()`
        compat/zlib: provide `deflateBound()` shim centrally
        git-compat-util: move include of "compat/zlib.h" into "git-zlib.h"
        compat: introduce new "zlib.h" header
        git-compat-util: drop `z_const` define
        compat: drop `uncompress2()` compatibility shim
  $ 
  
  $ cat GIT-VERSION
  v2.48.0
  $ 
 
An additional wrinkle on cygwin is that, after having built git (or more
precisely git-version.exe) then 'make sparse' and 'make hdr-check' spew to
stderr, like:

  /home/ramsay/git/GIT-VERSION-GEN: line 27: warning: command substitution: \
    ignored null byte in input 

which is caused by GIT-VERSION, git-version and git-version.exe being 'seen' as the same file on cygwin (case insensitive filessytem + cygwin 'magic'):

  $ ls -l GIT-VERSION
  -rwxr-xr-x 140 ramsay None 21M Jan 14 00:05 GIT-VERSION*
  $ 
 
  $ file GIT-VERSION
  GIT-VERSION: PE32+ executable (console) x86-64, for MS Windows, 19 sections
  $ 
 
  $ ls -l *version*
  -rwxr-xr-x 140 ramsay None  21M Jan 14 00:05 git-version.exe*
  -rw-r--r--   1 ramsay None 3.4K Jan 13 23:49 version.c
  -rw-r--r--   1 ramsay None  619 Jan 13 23:49 version.h
  -rw-r--r--   1 ramsay None   50 Jan 14 00:24 version.hcc
  -rw-r--r--   1 ramsay None  21K Jan 14 00:13 version.o
  -rw-r--r--   1 ramsay None    0 Jan 14 00:13 version.sp
  -rw-r--r--   1 ramsay None 5.4K Dec 24 14:03 versioncmp.c
  -rw-r--r--   1 ramsay None  118 Sep  8 21:39 versioncmp.h
  -rw-r--r--   1 ramsay None   53 Jan 14 00:24 versioncmp.hcc
  -rw-r--r--   1 ramsay None  15K Jan 13 23:59 versioncmp.o
  -rw-r--r--   1 ramsay None    0 Jan 14 00:13 versioncmp.sp
  -rw-r--r--   1 ramsay None  199 Jan 14 00:13 version-def.h
  -rw-r--r--   1 ramsay None  207 Dec 16 19:08 version-def.h.in
  $ 

I normally like to find solutions before posting problems to the list, but unfortunately I don't have time to look into this at the moment. Sorry!

Just a heads up.

ATB, Ramsay Jones

Ramsay Jones <ramsay@ramsayjones.plus.com> writes:

This came from semantic mismerge with Patrick's latest round of build updates. Usually I try to review very critically parts of a series that can negatively affect those users who do *not* use a new feature each topic tries to add or improve, and it means for this topic, if it breaks builds that do not involve meson, it would automatically qualify the topic to be ejected from the tree (side note: if a breakage is for those who use meson, as long as it is not as grave as running an equivalent of "sudo rm -fr /", I'll let it through), but this time because there was no textual conflict, the hardcoded GIT-VERSION unfortunately went through.

I however somehow thought that those who are building from a repository should get "git describe" name instead of the hardcoded fallback name in GIT-VERSION. At least, it is my understanding that is the intent of that series. Or are you building from a tarball extract?

I'd throw it back at Patrick, at least for tonight; the topic is more or less hot off the press and it should not be surprising it it needed immediate fix-up and reroll.

Thanks.

On Mon, Jan 13, 2025 at 05:49:59PM -0800, Junio C Hamano wrote:

Yup, found the issue. The next version of the Meson series will have it fixed. Thanks for the report!

Patrick

Hi Patrick, Junio,

On 14/01/2025 07:51, Patrick Steinhardt wrote:

[snip]

I suspect you already know, but I can confirm that tonight's 'seen' branch builds fine (Linux and Cygwin).

[Sorry for being so late, but I had to wait for about 3 hours for windoze update to stop making my laptop *totally* unusable! :( ]

Thanks!

ATB, Ramsay Jones