Re: [TopGit PATCH] tg-push: prevent multiple occurences of branches

[Bert Wesarg <bert.wesarg@googlemail.com>]

2009/5/27 Uwe Kleine-König <u.kleine-koenig@pengutronix.de>:
> Hi Bert,
Moin Uwe,

>
> On Tue, May 26, 2009 at 11:37:00PM +0200, Bert Wesarg wrote:
>> So, I finally tried tg-push and got a lot of warnings and errors:
>>
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>> warning, no base found top-bases/master
>>
>>
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/master is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/master
>> error: Ref refs/heads/bw/log is at a16df35cc7009b36f6f71717ae3d9a3dc29987da but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/heads/bw/log
>> error: Ref refs/top-bases/bw/log is at b725fc951a37854fc65945ac5bcab3bc61ccbd94 but expected 0000000000000000000000000000000000000000
>> error: failed to lock refs/top-bases/bw/log
>>
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] master -> master (failed to lock)
>>  ! [remote rejected] bw/log -> bw/log (failed to lock)
>>  ! [remote rejected] refs/top-bases/bw/log -> refs/top-bases/bw/log (failed to lock)
> hhm, does this happen because
>
>        git push $remote master master master master
Exactly.

>
> tries to lock master four times?  I'd think this is something that needs
> fixing in git-core.
Cc'd Junio. And Junio, you can stop reading now.

>
>> Also the result was ok this output is defintive not ok.
>>
>> To prevent multiple occurences of the same branchname we maintain these in a
>> file.
>>
>> In push_branch ref_exists was called, which is redundant because _dep_is_tgish
>> tells us exactly this.
> ... what about breaking this out?
By your command.

>
>> ---
>>  tg-push.sh |   48 +++++++++++++++++++++++++++++-------------------
>>  1 files changed, 29 insertions(+), 19 deletions(-)
>>
>> diff --git a/tg-push.sh b/tg-push.sh
>> index 8d09a02..c813927 100644
>> --- a/tg-push.sh
>> +++ b/tg-push.sh
>> @@ -45,31 +45,41 @@ for name in $branches; do
>>       ref_exists "$name" || die "detached HEAD? Can't push $name"
>>  done
>>
>> +_listfile="$(mktemp -t tg-push-listfile.XXXXXX)"
>> +trap "rm -f \"$_listfile\"" 0
>> +
>> +# prevent multiple occurrences of the same branch
>> +add_to_list()
>> +{
>> +     fgrep -qx "$1" "$_listfile" ||
>> +             echo "$1" >> "$_listfile"
>> +}
> This has complexity O(n**2) for n calls (I think).  Just writing then to
> the list and doing sort -u afterwards only has O(n log n) (for a
> reasonable implementation of sort).
Ahh, that takes a while, but yes. I thought I need to sort each time I
put a new branch in this file.

>
>>  push_branch()
>>  {
>>       # if so desired omit non tgish deps
>>       $tgish_deps_only && [ -z "$_dep_is_tgish" ] && return 0
>>
>> -     echo "$_dep"
> what about $tgish_deps_only?
Hmm, I think I have not changed the semantics here. I replace this
line with 'add_to_list "$_dep"'.

>> -     local base="top-bases/$_dep"
>> -     if ref_exists "$base"; then
>> -             echo "top-bases/$_dep"
>> -     else
>> -             echo "warning, no base found $base" 1>&2
>> -     fi
>> +     add_to_list "$_dep"
>> +     [ -z "$_dep_is_tgish" ] ||
>> +             add_to_list "top-bases/$_dep"
>
Bert

> Best regards
> Uwe