Re: [gentoo-dev] Removal of CVS headers

[Lars Wendler]

On Sun, 26 Feb 2017 16:42:59 -0500 Mike Gilbert wrote:

>On Sun, Feb 26, 2017 at 4:16 PM, Lars Wendler
><polynomial-c@g.o> wrote:
>> On Sun, 26 Feb 2017 22:07:48 +0100 Ulrich Mueller wrote:
>>
>>>>>>>> On Sun, 26 Feb 2017, Robin H Johnson wrote:
>>>
>>>> The 2014-10-14 meeting did NOT specify what CVS headers were in
>>>> question, and it was later decided that this was $Header$, not
>>>> $Id$.
>>>
>>>When and by whom was that decided? The unanimous council decision was
>>>to remove "CVS headers" and the obvious understanding is that this
>>>includes all Header, Id, Date, and so on. Quoting the actual agenda
>>>item (which was submitted by another infra member, BTW):
>>>
>>># 3. CVS Headers
>>>#
>>># The hateful thing. We could supposedly somehow fill them in rsync
>>># but that's complex and very dangerous (think of all the broken
>>>patch # files currently in gx86). I think we should kill them.
>>>#
>>># And while at it, I think it'd be good to actually remove most of
>>># them from our files -- changing header templates and so on. While
>>># not strictly useful, it decreases the size of the repo a bit and
>>># avoids any future nightmares :).
>>>
>>>Then in the summary of the 2016-11-13 council meeting we have this:
>>>
>>>- Bug 579460 "please make repoman ignore a missing "# $Id$" header
>>>line":
>>>  Implemented in repoman-2.3.0, but not yet in stable. Once this is
>>>  done, CVS headers can be removed as per 2014-10-14 council
>>> decision.
>>>
>>>Note that it explicitly mentions $Id$, and until today nobody had
>>>raised any objections against its revoval.
>>
>> Well, here is my objection now. And it seems like nobody at the
>> council seems to have understood the intention of this $Id$ header in
>> git.
>> See https://git-scm.com/docs/gitattributes and search for "ident".
>>
>>>> This was also discussed in August 2015:
>>>> Subject: 'Infra plans regarding $Id$ - official answer...'
>>>> https://archives.gentoo.org/gentoo-dev/message/d01ce943a9f9404c454c26bdb7efdf0e
>>>
>>>> $Id$ is used by Git as well, and I was a strong advocate that
>>>> expansion of $Id$ should be ENABLED in the rsync exports, because
>>>> it allowed tracing what version of a file was actually in use.
>>>
>>>I don't see any expansion of $Id$ in rsync as of today, 18 months
>>>after the Git conversion. If it wasn't missed for 18 months, one can
>>>hardly claim that it would be an important feature.
>>>
>>>Also I suspect that it is too late to enable it now, because it will
>>>potentially break patches that have been added after the conversion
>>>to Git.
>>
>> There is no need to enable it by default. But it is a very nice way
>> to verify ebuild changes if being enabled locally on a git clone of
>> the tree. Ever since portage was migrated to git I had this line in
>> my .git/info/attributes file on my dev box:
>>
>> *.ebuild ident
>>
>> This is a very useful feature and should not be removed only because
>> council was told that it's a mere CVS migration cruft. It is not!
>
>If this is about keeping ebuilds in your overlay in sync, you could
>alternatively use the output of git ls-tree to keep track of changes
>using the same blob ids.

This is not the same. git's ident mechanism doesn't use the commit blob
hash but creates a hash from the file's content. And as I am not using
VCS for my overlay, git ls-tree is no useful replacement solution here.

>It may not be as immediately convenient for you, but it would
>ultimately be more reliable in the long run; there are already several
>ebuilds in the gentoo repo that are missing the $Id$ header. Even if
>we restore the repoman check, I'm sure they will creep in.

Which then could be a perfect opportunity for QA to finally do
something useful.

-- 
Lars Wendler
Gentoo package maintainer
GPG: 21CC CF02 4586 0A07 ED93  9F68 498F E765 960E 9B39

Attention! New gpg key! See
https://www.gentoofan.org/blog/index.php?/archives/9-New-gpg-keys.html