[oe] angstrom-2008.1, uclibc: git-native: sysroot-destdir 6, 4 GB big, since some executables are 100 MB each

[Paul Menzel]

Am Samstag, den 17.07.2010, 14:15 +0200 schrieb Koen Kooi:
> On 17-07-10 10:35, Paul Menzel wrote:
> > Am Samstag, den 17.07.2010, 01:03 +0200 schrieb Koen Kooi:
> >> On 16-07-10 17:23, Khem Raj wrote:
> >>> On Fri, Jul 16, 2010 at 2:03 AM, Paul Menzel
> >>> <paulepanter at users.sourceforge.net> wrote:
> > 
> >>>> Am Donnerstag, den 15.07.2010, 19:55 +0200 schrieb Paul Menzel:
> >>>>
> >>>>> I ran out of space and noticed
> >>>>>
> >>>>>     /angstrom-dev/work/i686-linux/git-native-1.7.0.2-r4
> >>>>>
> >>>>> is 8,1G big.
> >>>>>
> >>>>> Some of the executable files in
> >>>>>
> >>>>>     /angstrom-dev/work/i686-linux/git-native-1.7.0.2-r4/sysroot-destdir/angstrom-dev/sysroots/i686-linux/usr/libexec/git-core/
> >>>>>
> >>>>> are 101 MB big.
> >>>
> >>> its most probably the -g3/-ggdb3 bloat.
> >>
> >> No, it's git not creating (hard)links but copies of the same app. Run an
> >> md5sum on the binaries, they will all match. I suggest you stop
> >> spreading FUD about -ggdb3, it is not being appriciated.
> > 
> > I do not know anything about the technical background. I can only report
> > that removing `-ggdb3` fixed it for me.
> 
> Are you saying that removing -ggdb3 made git create proper (hard)links?

No, I just said, that after doing as Khem suggested the available space
was big enough.

Anyway, I tried to investigate a little and could not find anything
related to (hard)link problems. But probably I overlooked something. I
head refresh my memory about hard and soft links [1].

1. As you suggested in your reply I ran `md5sum` on the binaries. They
do indeed match on the binaries I posted, but also the inode matches. It
looks like that is way it should be.

        $ echo "Test" > test
        $ ln test test.link
        $ md5sum test*
        2205e48de5f93c784733ffcca841d2b5  test
        2205e48de5f93c784733ffcca841d2b5  test.link
        $ ls -1i test*
        3653430 test
        3653430 test.link

So I think the hardlinks were created properly. I did not think of this
when reporting the problem so I did not run `ls -i` on the listed files
and I did not check with `du`.

        /angstrom-dev/work/i686-linux/git-native-1.7.1.1-r0/image/angstrom-dev/sysroots/i686-linux/usr/libexec/git-core $ md5sum * | sort
        […]
        82c05218d9634923e301a9463e8f9c0d  git-shell
        85edbd7803242b4895ed33b4fcfa3622  git-remote-ftp
        85edbd7803242b4895ed33b4fcfa3622  git-remote-ftps
        85edbd7803242b4895ed33b4fcfa3622  git-remote-http
        85edbd7803242b4895ed33b4fcfa3622  git-remote-https
        89a6ef52f4b7c336fb5fc0601edfd95f  git-mergetool
        9a7f760ecef6e64f370c907d61ad0ffb  git-svn
        9b57eb61ce668f2105a61a467f4d6d9b  git-difftool
        ac3472a81bc59a69119b0d027c3f617b  git-merge-one-file
        b14a5bba57d0ae0ccfdb0fceac0eb4ec  git-merge-octopus
        b6899828bc756c5d30398d49dce91deb  git-cvsexportcommit
        b775de2d21923f5d08e0bdcf3f9d31f6  git-mergetool--lib
        bd786a897bdc214bd5d89f51ad79f152  git-instaweb
        c8466908a274e211a43d287e510f2f19  git-merge-resolve
        ce109bd27d9404980bd82ddb529eca90  git-send-email
        d3da496bd962246727ceaabc1d4b1bdd  git-sh-setup
        d46c9b755b712bf3c27521b3c6e5fd30  git-relink
        d4b577d9a69d9af883de29d6af4f04c5  git
        d4b577d9a69d9af883de29d6af4f04c5  git-add
        d4b577d9a69d9af883de29d6af4f04c5  git-annotate
        d4b577d9a69d9af883de29d6af4f04c5  git-apply
        […]

        /angstrom-dev/work/i686-linux/git-native-1.7.1.1-r0/image/angstrom-dev/sysroots/i686-linux/usr/libexec/git-core $ ls -1i
        [ … ordered manually to match md5sum output]
        3691063 git-shell
        3691070 git-remote-ftp
        3691070 git-remote-ftps
        3691070 git-remote-http
        3691070 git-remote-https
        3691082 git-mergetool
        3691100 git-svn
        3691093 git-difftool
        3691080 git-merge-one-file
        3691079 git-merge-octopus
        3691095 git-cvsexportcommit
        3691102 git-mergetool--lib
        3691101 git-instaweb
        3691081 git-merge-resolve
        3691099 git-send-email
        3691104 git-sh-setup
        3691098 git-relink
        3691105 git
        […]

What did I miss? Were you seeing something different?

2. You seem to have investigated this problem before and committed a fix
for git [2].

    git 1.7.0.2: fix hardlinking problems as reported by Eric Brombaugh

I could not find Eric’s report on this list or the Web. So I have to
test without your patch. But `git-native` does not seem to use your
patch, so I guess the behavior should be the one without your patch.

Could you please tell me if my observations are correct.


Thanks,

Paul


[1] info coreutils "ln invocation"
[2] http://git.openembedded.org/cgit.cgi/openembedded/commit/?id=84de51631e1e7bb76141332807ed2ba9e8df02ae
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://lists.openembedded.org/pipermail/openembedded-devel/attachments/20100718/7e596934/attachment-0002.sig>