webrev.ksh inserts full commit history of a file
Volker Simonis
volker.simonis at gmail.com
Wed Sep 16 17:29:43 UTC 2015
Just for reference: the following bug already exists for this misbehaviour:
Mercurial 3.4 breaks behavior of webrev -r
https://bugs.openjdk.java.net/browse/CODETOOLS-7901466
On Tue, Sep 15, 2015 at 3:59 PM, Attila Szegedi
<attila.szegedi at oracle.com> wrote:
> Thanks both Volker and Daniel for responding.
>
> Apparently, the --follow option is causing the trouble (as used in `hg log …` command within the comments_from_mercurial() function. I forced webrev to print how it invokes hg, and it was sending it e.g.:
>
> hg log --rev 1507 --rev 1508 --follow --template 'rev {rev} : {desc}\n' src/jdk/nashorn/internal/codegen/AssignSymbols.java
>
> And this indeed creates the full history. If I remove “--follow” from the command line, I get the expected output. I now have Mercurial 3.5.1 (yep, I know, bleeding edge…). For now, I just removed “--follow” from my local copy of webrev.ksh, this seems like it’ll do for me.
>
> Thanks to everyone who offered to help!
>
> Attila.
>
>> On Sep 15, 2015, at 2:47 PM, Volker Simonis <volker.simonis at gmail.com> wrote:
>>
>> The thing I noticed in your bad examples is that in the you compare
>> against "ssh://hg.openjdk.java.net/jdk9/dev/nashorn". I'm not sure if
>> this is a problem (or maybe only a problem if you are sitting behind a
>> firewall). I don't have 'default-push' defined in my .hg/hgrc files so
>> webrev.ksh is always using the 'default' entry (which is a http-URL)
>> and I never saw these problems. Webrev.ksh first reads 'default-push'
>> and only if this isn't defined it reads the 'default' path. Maybe you
>> can give it a quick try and use "-p http://hg.open..." to see if this
>> helps?
>>
>> Also, which version of webrev.ksh are you using, the latest one?
>>
>> On Tue, Sep 15, 2015 at 1:32 PM, Attila Szegedi
>> <attila.szegedi at oracle.com> wrote:
>>> I knew about -r and have used it in the past; unfortunately it didn’t help. Even doing “webrev.ksh -N -r qparent” gives me the wrong results (still a full commit history).
>>>
>>> Attila.
>>>
>>>> On Sep 14, 2015, at 9:24 AM, Volker Simonis <volker.simonis at gmail.com> wrote:
>>>>
>>>> Hi Attila,
>>>>
>>>> you can use '-r rev' to compare against a specific revision. I use it
>>>> together with Mercurial Queues if I have changes from my queue pushed
>>>> but only want a webrev of the top-most change. In that case I do
>>>> "webrev.ksh -r qbase".
>>>>
>>>> Also you can use '-N' to prevent webrev.ksh doing 'hg outgoing' and
>>>> instead producing a webrev of local changes only (i.e. 'hg status').
>>>>
>>>> Regards,
>>>> Volker
>>>>
>>>>
>>>> On Fri, Sep 11, 2015 at 2:41 PM, Attila Szegedi
>>>> <attila.szegedi at oracle.com> wrote:
>>>>> Recently I noticed webrev.ksh started including the full commit history of files into the generated webrevs. E.g. see http://cr.openjdk.java.net/~attila/8135262/webrev.jdk9/ <http://cr.openjdk.java.net/~attila/8135262/webrev.jdk9/>
>>>>>
>>>>> I have seen some other people recently posting webrevs suffering from the same problem, e.g. <http://cr.openjdk.java.net/~aw/8134873/webrev.01/ <http://cr.openjdk.java.net/~aw/8134873/webrev.01/>>
>>>>>
>>>>> One thing I did recently was reinstall all my MacPorts as part of upgrading them for OS X 10.10, so my ports-provided Mercurial now identifies itself as “3.4.99” in port list and as “3.5-rc+12-a74e9806d17d” with “hg —version”. Not sure if that matters. If anybody has an insight into it, please let me know.
>>>>>
>>>>> Thanks,
>>>>> Attila.
>>>>>
>>>>>
>>>>>
>>>>>
>>>
>
More information about the code-tools-dev
mailing list