Too many large binary files in git history about clasp HOT 11 CLOSED

Thanks jasom - that's sounds great.

from clasp.

https://gist.github.com/anonymous/488b81ad81838291eb33

See the comment at the bottom for how to locally apply the changes

I can walk you through the steps of getting those changes pushed out to github on irc tomorrow. You're on an earlier timezone than me, so you might be able to locally apply and test it before I'm up.

from clasp.

I just wanted to make a quick comment here. I haven't actually gone through the clasp repos history carefully. However, from what someone said, it seems likely that the stuff that is now in externals is in the history of the clasp repos too, though not currently in head. I think those files should be removed as well. There is no point keeping them. You (Jason) said something about jpg and pdf files in there. Most likely those should go too. There are special circumstances under which is makes sense to keep jpg and pdf files (e.g. scans) under version control, but in most cases it does not make sense. And I don't think those special cases apply here.

There are two possibilities:

(a) the jpg/pdf files belong to some external dependency, in which case they should be removed (see above)

(b) they belong to clasp itself, in which case they should really be generated from source. Though I suspect this is not the case for most (all?) the files in question.

I think only text files pertaining to clasp itself should be retained as far as possible.
Dependencies which clasp is actually including (as opposed to linking against) will also need to stay, I guess. Thank you for your attention.

from clasp.

files in #41 can probably be unpersoned as well, though they're not generally huge.

from clasp.

For clasp:

git filter-branch --index-filter 'xargs git rm -rf --cached --ignore-unmatch < $HOME/filelist' -- --all

where $HOME/filelist is here

for clasp-externals:

git rev-parse --verify 89a444091fd1ca6e825f21a865e426a3eaf82ed8 >> .git/info/grafts && git filter-branch -- --all

from clasp.

Offtopic: This is why Perforce will beat Git for some time in games development. You need to check in movies and game assets.

from clasp.

@rptb1 : git-annex plus a NAS solves that problem, though it's perhaps not quite as nice as p4, git has other advantages.

from clasp.

git rev-parse --verify  cee291ccdcea092b732756dac0bd10bb9c901143 >> .git/info/grafts && git filter-branch -- --all

from clasp.

faheem mentioned that he thought the pack was still too big; I figured this was "good enough" but if someone wants to slog through, then I've pasted all remaining files larger than 10k in the pack, sorted by size in the pack:

https://gist.github.com/anonymous/55949a66ba5593accf1b

from clasp.

Here's a program that can generate the list of git files in a pack, requested by @fmitha https://github.com/jasom/gitfiles

Note that in order to use it you need to pack all lose objects (i.e. run git repack), as it can only find files that are in a pack.

from clasp.

I performed the cleanup as directed by jasom and faheem and now the repo is about 87MB when previously it was about 580MB. Everything seems to work fine.

Medals all around guys!

from clasp.