Re: [PATCH] Maintaince script for l10n files and commits

From: Jiang Xin <worldhello.net@gmail.com>
Date: Mar 9, 2012, 00:57 UTC
Message-ID: <CANYiYbGxsTZ1sDVf8e47w9butmN_XUgucJByfCdYfFSpNhzVYA@mail.gmail.com>
In-Reply-To: <7v399iddw3.fsf@alter.siamese.dyndns.org>

2012/3/9 Junio C Hamano <gitster@pobox.com>:

Show 18 quoted lines

> That's quite a style deviation from out norm in the commit log
> messages, don't you think (see "git log --no-merges -100", for
> example)?  State the problem you are attempting to solve first, and
> explain the solution to the problem, in separate paragraphs for
> readability, perhaps like this:
>
>        There are routine tasks translators need to perform that can
>        be automated.
>
>        Help them to
>
>         (1) initialize or update the message files;
>         (2) check errors in the message files they edited;
>         (3) check errors in their commits; and
>         (4) review recent updates to the message template file
>             they base their translations on.
>
>        by adding a helper script.

Thank you for provide better commit log, and I learn a lot from it. Writing in English still a big issue for me, and also obstacles to many l10n contributors. So your decision that l10n contributors can write commit logs in native language is very helpful, yet it has potential risk about wrong character encodings in commit log. So I need to write a helper for l10n team leaders, especially for myself, to detect bad commit log, because nobody knows all languages and encodings. Hacks outside of "po/" directory should be checked regularly also.

Show 6 quoted lines

>> @@ -0,0 +1,271 @@
>> +#!/bin/bash
>
> Is there any bash-ism in this script?  Otherwise please start this
> with "#!/bin/sh" to allow people who do not use bash to get involved
> in the project.

There are several regex match expressions written in bash style, which is not dash compatible. I will try to use grep and sed instead.

Show 10 quoted lines

> Will we later regret that we didn't give a command word for this
> one?  Two common sources of such risks are:
>
>  (1) it turns out XX.po matches the pattern we would want to use as
>     a command; and
>
>  (2) it turns out "init/update" is not the most often used action.
>
> I do not think (1) is likely. I do not think anybody can decide
> about (2) at this point yet.

The style of arguments comes from previous Rakefile implementation. I will change XX.po as a alias subcommand of init/update.

-- 
Jiang xin