On Feb 20, 2009, at 11:58 AM, Daniel Shahaf wrote:
> Dongsheng Song wrote on Thu, 19 Feb 2009 at 09:42 +0800:
>> The current diff very large and hard to read or review(e.g. r35966):
>> % svn diff -c 35966 | diffstat
>> es.po | 5255 +++++++++++++++++++++++++++++++++++++
>> +----------------------------
>> 1 file changed, 3047 insertions(+), 2208 deletions(-)
>>
>> If we strip location:
>> % diff es-r31547.po.no es-r35966.po.no | diffstat
>> unknown | 1804 ++++++++++++++++++++++++++++++++++++++++++++
>> +-------------------
>> 1 file changed, 1280 insertions(+), 524 deletions(-)
>>
>> As we can see, the diff much smaller(5255 => 1804) and clear.
>>
>> So I suggest we sort by file, and strip string locations before
>> commit po file:
>>
>> msgcat --width=80 --sort-by-file -o tmp.po xx.po
>> msgcat --width=80 --no-location -o xx.po tmp.po
>> rm tmp.po
>>
>
> Sounds sane (I don't use the translations), but would it be possible
> to
> still support the use case of "grep the *.po files for an error
> message
> to see what file.c:line in the source code it comes from"?
Whoa. Interesting concept, but what's the use case here?
-Hyrum
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1200845
Received on 2009-02-20 23:22:45 CET