Re: [PATCH v2] Use {git, hg, svn} commit id if available for xen_changeset

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 08/05/13 12:26, Ian Campbell wrote:
> On Wed, 2013-05-08 at 12:17 +0100, Andrew Cooper wrote:
>> On 08/05/13 12:05, Ian Campbell wrote:
>>> On Wed, 2013-05-08 at 06:42 +0100, Marek Marczykowski wrote:
>>>> As Xen uses git as primary repository, get git commit id for
>>>> xen_changeset info.
>>> I applied this to an hg repo and:
>>> $ chmod +x xen/tools/scmversion 
>>> $ rm xen/include/xen/compile.h
>>> $ make -C xen include/xen/compile.h
>>> make: Entering directory `/local/scratch/ianc/devel/xen-unstable.hg/xen'
>>>  __  __            _  _    _____                    _        _     _      
>>>  \ \/ /___ _ __   | || |  |___ /    _   _ _ __  ___| |_ __ _| |__ | | ___ 
>>>   \  // _ \ '_ \  | || |_   |_ \ __| | | | '_ \/ __| __/ _` | '_ \| |/ _ \
>>>   /  \  __/ | | | |__   _| ___) |__| |_| | | | \__ \ || (_| | |_) | |  __/
>>>  /_/\_\___|_| |_|    |_|(_)____/    \__,_|_| |_|___/\__\__,_|_.__/|_|\___|
>>>                                                                           
>>> make: Leaving directory `/local/scratch/ianc/devel/xen-unstable.hg/xen'
>>> $ grep CHANGESET xen/include/xen/compile.h
>>> #define XEN_CHANGESET		" 2>/dev/null | tr -d n)"
>>>
>>> Same seems to happen for a git repo too so its not VCS specific.
>>>
>>> It also happens if I do an actual build rather than forcing things as
>>> above.
>>>
>>> Running scmversion by hand appears to produce good output.
>>>
>>> Indentation in svmversion seems a bit off though, e.g. 
>>> +        printf '%s %s%s' "$date" git: $head
>>> +
>>> +               # Is this git on svn?
>>> +               if git config --get svn-remote.svn.url >/dev/null; then
>>> +                       printf -- 'svn:%s' "`git svn find-rev $head`"
>>> +               fi
>>>
>>> Seems to be tab vs space related, the printf is space indented and the
>>> rest is tab indented.
>>>
>>> Ian.
>> Can I please re-request a merge of this change (which in itself is good
>> and fine) with
>>
>> http://lists.xen.org/archives/html/xen-devel/2012-07/msg01886.html
> I see that Ian J had a comment, was that addressed?

It was addressed in so far as I stated that it didn't get executed
multiple times.  The archives seem to have broken the thread of
conversation, which had an extra back-and-forth on it.

>
>> The reason is that the changeset information is *still broken* for an
>> out-of-scm builds such as RPMs and DEBs.
>>
>> I will happily rebase the above path on top of this one, if it is accepted.
> This patch includes handling of a ".scmversion" file in the tarball.
> I've no idea how to make git archive generate that but perhaps it would
> be possible for it to be created as part of our tarball generation
> process (Ian J CCd) and/or by RPM/DEB packaging.
>
> I did think git archive automatically added a .gitarchive file to the
> output, but my version doesn't seem to do that.
>
> If you want to address Ian's concern and rebase that may still be
> worthwhile.
>
> Ian.
>

A sample .hg_archival.txt is:

repo: 6ed20573be52baab95207f52a8d28084f3a1967c
node: 3a180a46d747dddf77e9bc987f216aaeb84adea7
branch: default
latesttag: v2.3.0
latesttagdistance: 45

Which is sadly not enough to recreate the <rev>:<shorthash> changeset
string for HG.


For the background of this patch, XenServer uses a patchqueue. I have
specifically listed the base and pq versions separately, as the
changeset information for an `hg archive` of a base repo with a pushed
queue is functionally useless after either of the base or patch queue
gets changed.

~Andrew