>> * win_build/boinc_installshield.zip is the biggest space-eater in the 
>> repository at the moment. It's even corrupted, possibly because of EOL 
>> conversions.
>
>Rom, your call.

Go ahead and delete it, I don't know anything about it.

----- Rom

-----Original Message-----
From: [email protected] 
[mailto:[email protected]] On Behalf Of Oliver Bock
Sent: Thursday, July 19, 2012 6:17 AM
To: [email protected]; David Anderson (BOINC)
Cc: [email protected]
Subject: Re: [boinc_dev] Git Migration Status

On 7/19/12 5:53 , Nicolás Alvarez wrote:
> * Not much change in the openness and transparency of the migration.
> Was the conversion Oliver's one-man job?

Effectively, yes.

> If tracking down names for
> the SVN author-mapping was so hard, why did nobody ask for help in the 
> lists?

It's up to David and Rom to decide that. We kept the mapping in a private list 
for the time being as it contains e-mail addresses. IMHO I could certainly put 
it up on Trac since those addresses are/will be public as soon as we publish 
the git repo anyway.

David, your call.

> Where are the conversion and cleanup scripts, so that people can 
> contribute?

http://boinc.berkeley.edu/trac/wiki/GitMigration

> What tool was used for the conversion? (does it really take weeks to 
> run??)

git-svn and git (see link above)

> * In both boinc.git and boinccleaned.git, all the history before
> r12316 is duplicated. In some places even 3x or 4x. Some of these 
> duplicated chains of commits contain the openssl etc. binary 
> libraries. Apart from being a problem in itself, it also makes it 
> harder for me to review the history and possibly find other problems.
> http://wstaw.org/m/2012/07/19/plasma-desktopW10725.png

I'll look into that...

> * SVN usernames were replaced with full name + email address only for 
> the git commit Author. The Committer field still shows the SVN 
> username and repository UUID.

This is intentional, to have a (somewhat hidden) reference to the original info.

> * I'm pretty sure boincadmin is David Anderson, just like boincadm.
> It's currently mapped to "BOINC Administrator".

That's David's decision to make (he reviewed the mapping).

> * win_build/boinc_installshield.zip is the biggest space-eater in the 
> repository at the moment. It's even corrupted, possibly because of EOL 
> conversions.

Rom, your call.

> * boinc_core_release_5_9_9 is "hanging": doesn't branch off the main 
> history branch. In addition, it still contains openssl etc. compiled 
> binaries.

Please note that there's no 5.9 branch, just the 5.9.x tags in BOINC:
http://boinc.berkeley.edu/trac/browser/boinc/branches

The tag's not "hanging" since it got its correct parent:

commit in boinccleaned.git: 97ecbabd (SVN r12591) parent in boinccleaned.git: 
b917e356 (SVN r12584)

This reflects:
http://boinc.berkeley.edu/trac/log/boinc/tags/boinc_core_release_5_9_9

However, it got a second parent f8d47e0a that shows that the tagging in SVN had 
been tried/done a number of times, resulting in revisions 12585-12590. Those 
are the culprits:

http://boinc.berkeley.edu/trac/changeset/12585/boinc
http://boinc.berkeley.edu/trac/changeset/12586/boinc
http://boinc.berkeley.edu/trac/changeset/12587/boinc
http://boinc.berkeley.edu/trac/changeset/12588/boinc
http://boinc.berkeley.edu/trac/changeset/12589/boinc
http://boinc.berkeley.edu/trac/changeset/12590/boinc

There's nothing wrong with the migration here, it just reflects SVN's style of 
tagging/revisioning when a tag gets "reverted" and done again.
r12591 was the final attempt and that's mapped correctly in git.

> * The following work branches were not converted. There's only the 
> commits corresponding to the merges back to trunk, often doing massive 
> changes.

I asked the project team about the branches to be converted. Workspaces haven't 
been among those. David?

> * There's a lot of issues that are actually caused by bad usage of SVN 
> rather than a bad conversion. Even while on SVN, *please*:
>   - forget about checkin_notes; use the commit message
>   - don't do unrelated changes in a single commit (I can cite 
> thousands of examples)
>   - write descriptive commit messages ("${feature} stuff" and "Quick 
> Updates" aren't)
>   - don't list changed files in the commit message (Rom), it's 
> redundant since the VCS software shows that already.

Already discussed a number of times and I fully agree with you.


Nicolás, why don't you just go ahead, do it yourself and post the results? You 
can do this while we continue to have the discussion here and it should take 
you no time at all, right? I really don't care who does the migration and I'm 
not after any credit for doing this. I simply like to see BOINC migrated to git 
(as most of us do) so I stepped forward. You can take over at any time as long 
as you get things done.


Oliver

_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to