On 10 May 2011 11:04, Alex Blewitt
<alex.blewitt@xxxxxxxxx> wrote:
I think it makes sense. Yes, a single repository makes it easier and at 150Mb or so isn't the end of the world. Plus you only need to do it once :-)
It's about 250M for a faithful full-history clone:
(cvsimport is really really bad at producing correct conversion...). This will come down a fair bit by deleting 'old'. Checkout size:
bash:jamesb:xl-cbga-20:32971> du -sh *|grep M
13M build
1.2M c99
2.6M codan
3.2M contrib
43M core
9.1M debug
19M doc
6.1M dsf
5.4M dsf-gdb
56M edc
4.3M lrparser
1.2M memory
85M old
2.1M releng
1.5M upc
2.4M xlc
Those who just want to peek at the source can do a shallow clone anyway. (git clone --depth 1 will effectively lose history; it also means you won't be able to push back to the repository as well)
This could be useful for contributors. As we won't be allowed to use gerrit initially, all patches from contributors still need to go through bugzilla and be explicitly commited by a committer on their behalf. I.e. No pulling changes from non-committer friends...
Presumably --depth 1 prevents all further repo interactions from working: i.e. fetch / pull / push, so committers shouldn't use this.
I'm assuming that there's no inadvertently committed binaries etc? Is it worth doing a scan of the repo for known extensions (exe, tar, o, PDF?) to check for content that shouldn't be there?
We do have a few exes, not many. Mostly in the tests for the symbol readers. As a rule there's not much derived content to speak of in the repo.
Cheers,
James
