svn and git mirror status

hi everybody,

just wanted to send out a note, that the svn and git
mirrors of the LAMMPS are currently out of sync
due to hardware updates on the Sandia side of
the (complicated!) synchronization process.

we are working on resolving these issues.
as a consequence the svn and git mirror repositories
have not been updated for about a week, so please
don't be surprised to see them lagging behind and
missing the last few patches.

i'll post a follow-up as soon as we are back to normal.

thanks for your patience,
     axel

hi everybody,

the svn and git mirrors of the LAMMPS svn repository
at sandia are now linked to the new machine doing
the mirroring to the "server" at temple.

the update should be transparent to anybody updating
their LAMMPS version through svn or git.

please let me know, if there are any problems.

thanks,
    axel.

Axel,

I'm having a problem pulling from the git repository. My last successful "git pull" was on Oct 3, and since then I've made a few commits to my own local branch. But today (since the server is running again) I tried to pull and got this:

$ git pull
error: Unable to find 79df7ef875856ada56bba7c7f415c4343c8b1407 under http://git.icms.temple.edu/lammps-ro.git
Cannot obtain needed blob 79df7ef875856ada56bba7c7f415c4343c8b1407
while processing commit 860bca7d5e5d4d545fc43a3ef776e0bd604b35e9.
error: Fetch failed.

I don't know what that error means, or if the problem is on my end or the server or what. I was able to clone a fresh copy into a new directory successfully, but I would like to preserve my branch if possible.

Could you look into this? Thanks a bunch.

Axel,

hi sam,

I'm having a problem pulling from the git repository. My last successful
"git pull" was on Oct 3, and since then I've made a few commits to my own
local branch. But today (since the server is running again) I tried to pull
and got this:

$ git pull
error: Unable to find 79df7ef875856ada56bba7c7f415c4343c8b1407 under
http://git.icms.temple.edu/lammps-ro.git
Cannot obtain needed blob 79df7ef875856ada56bba7c7f415c4343c8b1407
while processing commit 860bca7d5e5d4d545fc43a3ef776e0bd604b35e9.
error: Fetch failed.

I don't know what that error means, or if the problem is on my end or the
server or what. I was able to clone a fresh copy into a new directory
successfully, but I would like to preserve my branch if possible.

you should be able to do a "local" push into the freshly cloned
directory, but that is beside the point. this kind of error should
not happen.

Could you look into this? Thanks a bunch.

this must be due to an additional package i had to configure
into the apache server to counter a bad vulnerability to
malformed HTTP requests.

it must block requests outside of what it is supposed to block.
can you tell me exactly the date and time (incl. time zone :wink: )
of your last pull request, so that i can more easily track them
down in the log files?

thanks,
    axel.

sam,

i don't see any offending error messages in the logs other than
refusals to serve files that don't exist.

i've manually rerun git-gc to clean up and recompress the
repository. now all "blobs" should be in a single "pack".

can you please try to pull again?

thanks,
    axel.

Axel,

I just tried to pull again and it worked this time. With some merge conflicts, but oh well. :slight_smile:

Thanks for your help.

hi sam,

Axel,

I just tried to pull again and it worked this time. With some merge

excellent. not sure what failed. if this happens more often,
i should probably run the garbage collection during the
automatic nightly update.

conflicts, but oh well. :slight_smile:

yeah, there are some (small) changes to many files due
to multi-threading support (you have to blame me for it).

i am very close to have steve releasing a much improved
variant in the next few days. for several styles and particularly
for inhomogeneous systems, multi-threading has become
quite competitive to MPI parallelization. e.g. folks using
AIREBO for CNT systems, should be able to run much
faster due to improvements in the pair style and better
load balance due to parallelizing over particles and not
domains.

Thanks for your help.

no problem. i am happy that we are back in business
and that it there are not no unsurmountable problems
despite have to rebuild the mirror repository from scratch.

cheers,
   axel.