On Mon, 16 Apr 2007, Martin S. Weber wrote:

> Hi Bernd,
>
> On Mon, Apr 16, 2007 at 06:49:24PM +0200, Bernd Ernesti wrote:
>> On Mon, Apr 16, 2007 at 01:34:01PM +0200, Martin S. Weber wrote:
>> [..]
>>
>>> But I shouldn't be offering patches for situations I don't fully understand.
>>> So: Why is e.g. freeciv-share ending up in "Failed" and e.g. curl isn't ??
>>
>> There is also a posibility that a 'make clean' after a 'make installed' can fail.
>> I know that at least doxygen, ezm3 or thunderbird fails while using pkg_chk
>> as a non root user and installed with sudo (SU_CMD).
>
> Well, the problem is that given the code of pkg_chk and the output I am
> witnessing that there is no way something that outputs "installed in previous
> stage" will end up in the failed-list as the deciding point (existance of the
> respective directory in the pkg-db) *cannot* be false (else there would be
> no output of "installed in previous stage").
>
> Which is why it surprises me that there are some packages which output "installed
> in previous stage" and end up in the install-list and others which output the
> same and end up in the fail-list.
>
> ... hmm makes me wonder where the FAIL variable gets reset which is checked to
> decide whether something should be in the fail-list or install-list...
>
> ... hmm it's reset in run_cmd which isn't reached if the pkg was built "in
> previous stage". Maybe pkg_install should do "FAIL= " at its head.

 	Thats exactly the case - good catch. Will update pkg_chk now

-- 
 		David/absolute       -- www.NetBSD.org: No hype required --