svn commit: r386285 - head/security/keychain

[Adam Weinberger]

> On 14 May, 2015, at 10:09, Bryan Drewery <bdrewery at FreeBSD.org> wrote:
> 
> On 5/14/2015 11:07 AM, Adam Weinberger wrote:
>>> On 14 May, 2015, at 10:03, Bryan Drewery <bdrewery at FreeBSD.org> wrote:
>>> 
>>> On 5/14/2015 11:00 AM, Adam Weinberger wrote:
>>>>> On 13 May, 2015, at 21:20, Philip M. Gollucci <pgollucci at p6m7g8.com> wrote:
>>>>> 
>>>>> Then it needs an option not an if exists()
>>>> 
>>>> Bryan, can you please consider reverting it, or giving me approval to revert it? It may fix the case when the port is built in poudriere without sshd in base, but it is a regression for everybody else.
>>>> 
>>>> # Adam
>>>> 
>>>> 
>>> 
>>> No. The way RUN_DEPENDS is supposed to work is what I said in the
>>> commit. These exists() checks are redundant and confuse the package
>>> dependencies.
>>> 
>>> How are you building? It's possible the tool you are using is
>>> incorrectly handling RUN_DEPENDS.
>> 
>> Ah! I see. It's portmaster that's doing the misbehaving. The problem is that openssh-portable still shows up in run-depends-list, and so portmaster decides that it is going to install openssh-portable first.
>> 
>> [root at apnoea /usr/ports/security/keychain] which ssh-agent
>> /usr/bin/ssh-agent
>> [root at apnoea /usr/ports/security/keychain] make run-depends-list
>> /usr/ports/security/openssh-portable
>> /usr/ports/shells/bash
>> 
> 
> Ok. I'll look at portmaster and try to fix it now. We really need these
> exists() to go away; it causes rebuilds on the package cluster for no
> reason. This specific case is not one, but we need them to generally go
> away.

Sorry about the noise, Bryan.

# Adam


-- 
Adam Weinberger
adamw at adamw.org
http://www.adamw.org