[CIG-SHORT] [petsc-maint] Git Version of PETSc does not compile

Satish Balay balay at mcs.anl.gov
Fri Feb 13 08:07:04 PST 2015


BTW: for our test - we just do a clone [if one doesn't exist] - and then do:

git fetch
git checkout origin/branchname

[i.e test in detached mode - so one doens't have to deal with dangling branches]

This way - the network traffic is minimal during the nightlybuilds.

You appear to do a 'git clone' each time - perhaps this is costly [wrt network traffic?]

Satish

On Fri, 13 Feb 2015, Satish Balay wrote:

> On Fri, 13 Feb 2015, Matthew Knepley wrote:
> 
> >     git clone --depth 1 -b knepley/pylith
> > https://bitbucket.org/petsc/petsc.git petsc-pylith
> > 
> > is messed up, but maybe Jed can tell us.
> 
> >>>>>>>>
> [balay at maverick ~]$ git version
> git version 2.1.0
> [balay at maverick ~]$ git clone --depth 1 -b knepley/pylith https://bitbucket.org/petsc/petsc.git petsc-pylith
> Cloning into 'petsc-pylith'...
> remote: Counting objects: 5495, done.
> remote: Compressing objects: 100 (5495/5495), 13.21 MiB | 1.64 MiB/s, done.
> Resolving deltas: 100[balay at maverick petsc-pylith (4836/4836), done.
> remote: Total 5495 (delta 1032), reused 2862 (delta 572)
> Receiving objects: 100 (1032/1032), done.
> Checking connectivity... done.
> [balay at maverick ~]$ cd petsc-pylith/
> [balay at maverick petsc-pylith]$ git describe
> fatal: No names found, cannot describe anything.
> [balay at maverick petsc-pylith]$ git log 
> commit 04fcc16b6fc1fef768ea84f574fa77a92547ade2
> Author: Matthew G. Knepley <knepley at gmail.com>
> Date:   Tue Feb 10 07:14:54 2015 -0600
> 
>     Merge branch 'knepley/fix-veclock-fftw' into knepley/pylith
>     
>     * knepley/fix-veclock-fftw:
>       DM: Fix more Vec lock issues
>       FFTW: Updated for vector locking
> [balay at maverick petsc-pylith]$ 
> <<<<<<<<<<<<<
> 
> So '--depth 1' just preserves only 1 commit - and not anything else from git history?
> 
> For one - perhaps we should use 'git fsck'? to do any checks - and not rely on 'git describe'?
> https://bitbucket.org/petsc/petsc/commits/6e709e5a83e3e6154cead97c7482e3df92822301
> 
> And secondly - for VERSION_GIT - if 'git describe' breaks - we should
> just do 'git log -1 --pretty=format:%H' - as we used to do before..
> 
> Satish
> 



More information about the CIG-SHORT mailing list