Hi Yann,

On Thu, 8 Dec 2016 10:41:29 +0100 Yann Leboulanger wrote:

> On 12/07/2016 11:37 PM, Andrey Gursky wrote:
> > Hi,
> >
> > a couple of days ago I noticed, that mercurial repository has gone. But
> > no replacement was setup. I haven't found any announcement here on the
> > list about the migration. Couldn't you switch it just into read-only
> > mode, which is already the only possible way to access the server for
> > everybody except developers?
> 
> I don't want ppl to think it's still maintained. So I prefer completly
> hide it.

OK, in this case the repositories could be renamed. 

> > The old trac has gone either. What do you think about an archive
> > read-only mode access?
> 
> Hard to make it read only. And why make it read only? (nearly) all is
> supposed to be on gitlab

Just in case there is something seems to be wrong and only look into
the original can clarify the situation.

> > Let's look at https://dev.gajim.org/gajim/gajim/issues/8390
> > It has been filed by anonymous. Can this be true? Did old trac allow
> > anonymous accounts (without at least partly anonymezed email reference)?
> 
> I didn't create an account for all 2500 contributors that filles a
> ticket. So yes we loose who created some tickets.

This is very unfortunately. But why are there so many accounts? I
believe it is due to the fact Gajim trac maintainers (were there many
or only you?) forced every user, wanting just to report an issue with
Gajim instead to take time to create an account, verify email, login.
Now I'm kindly ask you to be consequent and please continue to maintain
people contribution credits. It doesn't mean you would have to create
all these accounts. The only thing has to be done, is to prepend every
"anonymous" post with a corresponding original account name and email
(spam protected). This is very important, to encourage people to post
feedback to make Gajim better and get credited instead of showing,
that all they are just nobody anonymous (so why should they bother with
creating new accounts on gitlab and wasting time with filing bug reports?).

Theoretically, if the password hashing mechanism of trac and gitlab
were identical, I think accounts with old credentials could be created
and used without noticing any difference. How passwords are stored on
the old trac and new gitlab?

> > Looking further at the comment:
> > https://dev.gajim.org/gajim/gajim/issues/8390#note_137507
> > a changeset is mentioned:
> > In [changeset:"849a745fc6c17d18626b480d4ed7d3844be8280e" 16114:849a745fc6c1]
> > But I can't click on it to be forwarded to the actual commit. Hopefully,
> > you can fix that?
> 
> it's is now.

Thanks. (Though issues pages are returning 500 now.)

> > For example here https://dev.gajim.org/gajim/python-nbxmpp/issues/20
> > The issues and commits references are correctly linked.
> >
> > Let's look at https://dev.gajim.org/gajim/gajim/issues/8266
> > Again, an anonymous has supplied a patch. But the patch looks like to
> > be linked to Thilo Molitor's post. And in the commit
> > https://dev.gajim.org/gajim/gajim/commit/766bb508e323f849d
> > it seems the author is someone with nick gdr_gdr. But despite of the
> > commit comment "Fixes # 8266" the issue remained open. Moreover, due to
> > the space between # and 8266 gitlab cannot replace it with a proper
> > hyperlink. Maybe that is the reason for the stale status of the issue?
> 
> Maybe. Gitlab is currently very broken, so  I can't look at it. I'll do
> later.

Good, that this is not yet considered as final status.

> > To check such issues, a read-only archive mode access to old trac would
> > be helpful.
> 
> trac is still online on a hidden address as it's not read only.
> 
> > Now let's look how much space the old mercurial repository occupies:
> > $ du -sh .hg
> > 80M .hg
> >
> > Let's clone the new repository:
> > $ date
> > Tue Dec  6 19:45:46 CET 2016
> > $ git clone https://dev.gajim.org/gajim/gajim.git gajim.git
> > Cloning into 'gajim.git'...
> > remote: Counting objects: 93373, done.
> > remote: Compressing objects: 100% (27957/27957), done.
> > remote: Total 93373 (delta 63116), reused 93357 (delta 63105)
> > Receiving objects: 100% (93373/93373), 432.84 MiB | 5.52 MiB/s, done.
> > Resolving deltas: 100% (63116/63116), done.
> > Checking connectivity... done.
> > $ du -sh gajim.git/.git
> > 436M    gajim.git/.git
> > $
> 
> I know. I'm testing things.

Good. Thanks.

> try:
> git gc --aggressive --prune=now
> it reduces it to 59M

This is what I've already showed in detail in my previous mail:

> > That's really very much data. Please run git gc --aggressive on the
> > repository to avoid wasted traffic and doing this by everyone who
> > clones the repository:
> > $ du -sm .git; git gc --aggressive; du -sm .git
> > 436     .git
> > Counting objects: 93373, done.
> > Compressing objects: 100% (91048/91048), done.
> > Writing objects: 100% (93373/93373), done.
> > Total 93373 (delta 70902), reused 18994 (delta 0)
> > 35      .git
> > $
> >
> > 436 --> 35 MB, not bad (it took 2 min 25 sec on a 5 year old laptop)!
> > And efficiently than mercurial. Similar with gajim-plugins.
> >
> > It would be great, if you could enable git:// access to the repositories.
> 
> it is enabled. Create an accoun, import ssh  key and use git://

Almost all well known git servers, like kernel.org, repo.or.cz,
gnome.org, github.com,.. allow anonymous git:// access. Why should it
be so complicated with gitlab?

> > Those were my first impressions. Thanks for your attention,
>
> Thanks for your feedback. It's monthes I prepare the migration, it's a
> lot of work. I try to do my best.

Couldn't know, that you were working on this since all this time. Any
plan to blog on it? Have you used articles from others experience?

Regards,
Andrey
_______________________________________________
Gajim-devel mailing list
Gajim-devel@gajim.org
https://lists.gajim.org/cgi-bin/listinfo/gajim-devel

Reply via email to