[aur-general] avoid usage of fakeroot: always use of build() and package() (Was: application as a TU)

[Allan McRae]

hollunder at gmx.at wrote:
> On Sun, 22 Nov 2009 20:12:24 +1000
> Allan McRae <allan at archlinux.org> wrote:
> 
>> hollunder at gmx.at wrote:
>>> On Sun, 22 Nov 2009 16:17:36 +1000
>>> Allan McRae <allan at archlinux.org> wrote:
>>>
>>>> Ray Rashif wrote:
>>>>> 2009/11/22 Gerardo Exequiel Pozzi <vmlinuz386 at yahoo.com.ar>
>>>>>
>>>>>> Lukáš Jirkovský wrote:
>>>>>>> 2009/11/21 Gerardo Exequiel Pozzi <vmlinuz386 at yahoo.com.ar>:
>>>>>>>
>>>>>>>> hollunder at gmx.at wrote:
>>>>>>>>
>>>>>>>>> On Sat, 21 Nov 2009 14:24:58 -0300
>>>>>>>>> Gerardo Exequiel Pozzi <vmlinuz386 at yahoo.com.ar> wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> Andrea Scarpino wrote:
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>> Why do you use package() function when the package isn't a
>>>>>>>>>>> splitted package?
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>> Hello :)
>>>>>>>>>>
>>>>>>>>>> Using both build() and package() is not necessary condition
>>>>>>>>>> for use only with splitted packages, its avoid to use the
>>>>>>>>>> fakeroot on building process that is not needed in 99% of
>>>>>>>>>> packages.
>>>>>>>>>>
>>>>>>>>>> Good luck!
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> Sorry, but I consider the use of fakeroot a good thing, it
>>>>>>>>> helps to reveal errors while packaging/creating the PKGBUILD
>>>>>>>>> at least. Don't know why it should be avoided.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>> fakeroot make a table of function pointers for many file
>>>>>>>> manipulation calls, like open(), close(), chmod() and etc ->
>>>>>>>> overhead, slowdowns (small of course)
>>>>>>>> During the build process file perms are not necessary to be
>>>>>>>> "tracked", or at least in 99% of packages. Only during the
>>>>>>>> install process is only needed.
>>>>>>>>
>>>>>>>> If you have an example that breaks this, please let me know ;)
>>>>>>>>
>>>>>>>> --
>>>>>>>> Gerardo Exequiel Pozzi ( djgera )
>>>>>>>> http://www.djgera.com.ar
>>>>>>>> KeyID: 0x1B8C330D
>>>>>>>> Key fingerprint = 0CAA D5D4 CD85 4434 A219  76ED 39AB 221B 1B8C
>>>>>>>> 330D
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>> Nice feature! I didn't know that using package() avoids using of
>>>>>> fakeroot.
>>>>>>> Back to my point. There used to be a problem with compilation of
>>>>>>> amarok1 package from AUR only because of fakeroot and I guess
>>>>>>> that it would also help with building of mplayer (configure
>>>>>>> crashes under fakeroot environment and needs to be patched but
>>>>>>> maybe it was fixed meanwile alongside with amarok1 problem). So
>>>>>>> in some specific cases I can see the point of using separate
>>>>>>> package() even when the PKGBUILD builds only one package.
>>>>>>>
>>>>>>> best,
>>>>>>> Lukas
>>>>>>>
>>>>>>>
>>>>>> ;)
>>>>>>
>>>>>> Most reported crashes on bugtracker are because nvidia libgl,
>>>>>> that conflics with libfakeroot, both uses dlsym() (nvidia i
>>>>>> don't know why does this, fakeroot do this to fill a table of
>>>>>> pointer to functions), the result is a jump to NULL :P
>>>>>> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=516024#75
>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Gerardo Exequiel Pozzi ( djgera )
>>>>>> http://www.djgera.com.ar
>>>>>> KeyID: 0x1B8C330D
>>>>>> Key fingerprint = 0CAA D5D4 CD85 4434 A219  76ED 39AB 221B 1B8C
>>>>>> 330D
>>>>>>
>>>>>>
>>>>>>
>>>>> Hi Gerardo can you confirm that there are no (owner) permission
>>>>> issues when building without fakeroot (i.e all owned by user)? I
>>>>> don't see chown anywhere other than extract_sources().
>>>>>  
>>>> If your packaging is done properly, then there is no issues.
>>>> Always use "install" and not "cp".
>>>>
>>>> Allan
>>>>
>>> I've often seen cp used in makefiles, would this cause problems?
>>> Would you need to patch those makefiles?
>> If it breaks stuff, then you should tell upstream that they makefile
>> is crap.
> 
> So I guess this means it breaks?
> I usually do. You'd be surprised how many don't know about DESTDIR and
> use cp.

Probably not... makepkg sets a "sane" umask and the install is done as 
(fake)root.  So  you should be fine unless something is really screwed 
in the makefile.