[csw-maintainers] Fwd: Solaris package, was Re: Mercurial 1.3 released!

[rupert THURNER]

hi,

could it be possible that the type of providing the user homedir used
on the csw servers influence the test result of the mercurial package?
see the bottom of belows mail from mads.

i will try in the next weeks on a std filesystem as well with a std checkout.

rupert.

---------- Forwarded message ----------
From: Mads Kiilerich <mads at kiilerich.com>
Date: 2009/7/25
Subject: Re: Solaris package, was Re: Mercurial 1.3 released!
To: "rupert.thurner" <rupert.thurner at gmail.com>
Cc: mercurial at selenic.com, Trygve Laugstøl <trygvis at inamo.no>


rupert.thurner wrote, On 07/25/2009 11:52 AM:
>
> On Jul 5, 1:23 pm, Mads Kiilerich<mads at kiilerich.com>  wrote:
>
>>
>> rupert.thurner wrote, On 07/04/2009 08:13 PM:
>>
>>
>>>
>>> we tried to package it for solaris,http://opencsw.org. some test
>>> cases fail. some failures are because the test suite did not use /opt/
>>> csw/bin/gdiff, but "diff" from solaris. how do we specify the diff
>>> tool used, or the path to it?
>>>
>>
>> The testsuite uses whatever diff tool is in its path. So you could just
>> set PATH correctly before running the test suite.
>>
>>
>>>
>>> with some other tests i have no idea where the failure is coming from.
>>> what would be the best option to track this down? seehttp://pastebin.com/m32e3576f
>>> for the complete test result.
>>>
>>
>> test-repair-strip shouldn't rely on the cat error message. And the
>> warning in test-gpg should probably just be ignored. Patches has been
>> posted to -devel.
>>
>> Just ignore these tests for now.  You can temporarily remove their x bit
>> to avoid running them.
>>
>> That will make the failure list shorter and make it easier to spot:
>>
>>
>>>
>>> an excerpt:
>>>
>>
>>
>>>
>>> ERROR: test-transplant output changed
>>> --- Expected output
>>> +++ Test output
>>> @@ -145,20 +145,20 @@
>>>   % test filter
>>>   filtering
>>>   applying 17ab29e464c6
>>> -17ab29e464c6 transplanted to e9ffc54ea104
>>> +17ab29e464c6 transplanted to 88e81fba5b32
>>>   filtering
>>>
>>
>> This kind of errors looks worse. Do you consistently get exactly the
>> same errors?
>>
>
> i run the tests again with mercurial-1.3.1, see the result here:
> http://pastebin.com/m1a2f283f. especially
>

It seems like pastebin obscures the output slightly. Perhaps it would
be easier to read if it didn't use Python highlighting.

http://www.selenic.com/hg/index.cgi/rev/6e5e548452de (test-gpg:
suppress secure memory warning) should probably have been applied to
stable as well. Anyone?

Regarding test-repair-strip: One of the patches from
http://www.selenic.com/pipermail/mercurial-devel/2009-July/thread.html#13610
should be applied to stable.

You are still running in an environment without GNU diff. And
apparently also an unsupported version of cvs.

The rest of the problems seems to have something to do with something
not being like someone thinks they should be, perhaps forking or
caching or file systems.

Do you get the same failures when you run "make tests" manually from a
source checkout?