About zope and plone

[wen heping]

Yes, I agree with this PR that we should add -N to easy_install's
deinstall argument.

And shall we ask for a exp-run to test it ?

wen

2011/10/19 Ruslan Mahmatkhanov <cvs-src at yandex.ru>:
> Good day, gentlemen.
>
> Please tell what do you think about message bellow and this pr:
> http://bugs.freebsd.org/159962
>
> Message bellow is somewhat complements this pr. Please also note that Wen
> agree with this pr, but he is busy for this right now. Also note that this
> pr is a stopper wrt importing of new zope and plone.
>
> Thanks in advance.
>
> -------- Исходное сообщение --------
> Тема: Re: About zope and plone
> Дата: Sun, 16 Oct 2011 23:03:41 +0400
> От: Ruslan Mahmatkhanov <cvs-src at yandex.ru>
> Кому: wen heping <wenheping at gmail.com>
>
> wen heping wrote on 14.10.2011 13:15:
>
>> Today I tested devel/py-zope.minmax in Tinderbox, it has the same plist
>> error.
>> Now I need such a port exist in current FreeBSD portstree to be a demo
>> to python at .
>>
>> If python@ does not disagree I would ask portmgr@ for an exprun to test
>> it.
>> Then I would commit it.
>
> Hi Wen,
>
> i'm finally get it.
> While seeking an example for you i took some broken/outdated ports, that i
> just can't left in this state :), so i apologize for delay.
>
> You was right, it's a kind of upstream problem, but bsd.python.mk still
> contains a problem too, because this error still exists on deinstall of
> such silly ports (that may potentially exist in future).
>
> Here is how to reproduce:
> Pick any port, that
>  a) using setuptools for installing
>  b) has non-empty install_requires[] list in it's setup.py
>  c) does not define BUILD or RUN DEPENDS for deps, that are listed
>     in it's setup.py
>
> Ok, i wasn't able to find such port in the tree. But you can take, say,
> devel/py-daemon, remove BUILD/RUN_DEPENDS in it's Makefile and try to
> build it in tinderbox - it will builds and installs fine, but you'll get
> extra files installed on deinstall.
>
> The only port, that looked promising, was devel/py-Jinja. It defines
> this lines:
>
> PYEASYINSTALL_INSTALLARGS=      -N ${PYTHON_SITELIBDIR}/${PYEASYINSTALL_EGG}
> PYEASYINSTALL_UNINSTALLARGS=    -q -m -N
> ${PYTHON_SITELIBDIR}/${PYEASYINSTALL_EGG}
>
> Note adding the -N key on deinstall. But right now they are not needed,
> because this port packages successfully without them (it has empty
> install_requires[]).
>
> The problem with Products.MailHost, for example, is that if i add zope
> as dependency of this port, we well have cyclic deps and broken build.
> Resolving this issue with upstream will take much time (i know this
> because i already have 3 patches that hanging in zope/plone bugtrackers
> more than month).
>
> I can avoid this a different way, by defining
> PYEASYINSTALL_UNINSTALLARGS, but i believe that more correct and more
> easy is to define this -N in bsd.python.mk, because it's just two bytes
> in one file - and this is all, problem solved. Otherwise, i will be
> forced to patch many ports with this UNINSTALLARGS line, effectively
> littering the tree and making it hard to maintain. We already have it in
> PYEASYINSTALL_INSTALLARGS for some reason anyway, so why to not define it in
> UNINSTALLARGS too for symmetry?
>
> I hope this arguments are sufficient to convince portmgr@ and make this
> change.
>
> --
> Regards,
> Ruslan
>
> Tinderboxing kills... the drives.
>