Re: [PATCH] tg-push: Filter out plain SHA1s from being pushed.

["Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>]

Hello,

On Sat, Feb 27, 2010 at 11:09:08PM +0100, Thomas Schwinge wrote:
> Otherwise, problems arise when branches depend (via .topdeps) on SHA1s:
> 
>     $ tg push -r tmp t/____longjmp_chk_cleanup_linux_x86_64
>     fatal: 77c84aeb81808c3109665949448dba59965c391e cannot be resolved to branch.
>     fatal: The remote end hung up unexpectedly
> ---
> 
> Hello!
> 
> I was facing the following problem:
> 
>     thomas@dirichlet:~/tmp/source/glibc/git/glibc.hurd $ bash -v -x /usr/bin/tg push -r tmp t/____longjmp_chk_cleanup_linux_x86_64
>     [...]
>     ++ read _dep
>     ++ missing_deps=
>     ++ rm /tmp/tg-depsfile.nCfywI
>     ++ return 0
>     ++ xargs git push tmp
>     ++ sort -u /tmp/tg-push-listfile.sZeHIr
>     fatal: 77c84aeb81808c3109665949448dba59965c391e cannot be resolved to branch.
>     fatal: The remote end hung up unexpectedly
>     rm -f "/tmp/tg-push-listfile.sZeHIr"
>     + rm -f /tmp/tg-push-listfile.sZeHIr
> 
> That's the actual problem:
> 
>     thomas@dirichlet:~/tmp/source/glibc/git/glibc.hurd $ git push tmp 77c84aeb81808c3109665949448dba59965c391e
>     fatal: 77c84aeb81808c3109665949448dba59965c391e cannot be resolved to branch.
>     fatal: The remote end hung up unexpectedly
> 
> ... which is due to:
> 
>     $ git show t/____longjmp_chk_cleanup_linux_x86_64:.topdeps
>     77c84aeb81808c3109665949448dba59965c391e
> 
> All other parts (at least those I tested / got to use so far) work just
> fine when passing SHA1s.
> 
> Here is a patch to make this work for tg push, too.
> 
>     thomas@dirichlet:~/tmp/source/glibc/git/glibc.hurd $ tg push -r tmp t/____longjmp_chk_cleanup_linux_x86_64
>     Counting objects: 196207, done.
>     Compressing objects: 100% (32953/32953), done.
>     Writing objects: 100% (196207/196207), 66.45 MiB | 4047 KiB/s, done.
>     Total 196207 (delta 163379), reused 191592 (delta 159243)
>     To /media/Stalin/tmp/glibc
>      * [new branch]      t/____longjmp_chk_cleanup_linux_x86_64 -> t/____longjmp_chk_cleanup_linux_x86_64
>      * [new branch]      refs/top-bases/t/____longjmp_chk_cleanup_linux_x86_64 -> refs/top-bases/t/____longjmp_chk_cleanup_linux_x86_64
> 
> 
>  tg-push.sh |    5 ++++-
>  1 files changed, 4 insertions(+), 1 deletions(-)
> 
> diff --git a/tg-push.sh b/tg-push.sh
> index cd208a0..28cad81 100644
> --- a/tg-push.sh
> +++ b/tg-push.sh
> @@ -71,6 +71,9 @@ for name in $branches; do
>  	$recurse_deps && [ -n "$_dep_is_tgish" ] &&
>  		no_remotes=1 recurse_deps push_branch "$name"
>  
> +	# filter out plain SHA1s
>  	# remove multiple occurrences of the same branch
> -	sort -u "$_listfile" | xargs git push $dry_run "$remote"
> +	grep -vE '^[0-9a-f]{40}$' < "$_listfile" \
> +		| sort -u \
> +			| xargs git push $dry_run "$remote"
>  done
probably this is good enough.  Still I wonder if there is a better way
to detect if a given rev is a sha1 or a branch.

test "$(git rev-parse "$rev")" = "$rev"

comes to mind.  Anybody who gives names to branches that match
^[0-9a-f]{40}$ has other problems probably.

I think I will take it anyhow (and fix up the broken indention).

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |