pcp
[Top] [All Lists]

ps configure test (was Re: [pcp] Solaris build notes)

To: "Josef 'Jeff' Sipek" <jeffpc@xxxxxxxxxxxxxx>, Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Subject: ps configure test (was Re: [pcp] Solaris build notes)
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Sun, 28 Jun 2015 21:02:21 -0400 (EDT)
Cc: pcp <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <20150628151915.GC101529@xxxxxxxxxxxxxxxxxxxxxxxxx>
References: <474160992.25596336.1435198655884.JavaMail.zimbra@xxxxxxxxxx> <1945955780.25644422.1435200191659.JavaMail.zimbra@xxxxxxxxxx> <20150625133756.GB101450@xxxxxxxxxxxxxxxxxxxxxxxxx> <20150625161654.GF101450@xxxxxxxxxxxxxxxxxxxxxxxxx> <700307243.26392851.1435272506780.JavaMail.zimbra@xxxxxxxxxx> <20150628151915.GC101529@xxxxxxxxxxxxxxxxxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: xi9j+IRf2WExAST3yypnfQv+LU+RwQ==
Thread-topic: ps configure test (was Re: [pcp] Solaris build notes)

----- Original Message -----
> ...
> Ok, so I started working on setting up a host.  Since I have zones at my
> disposal, I decided to use a zone instead of a full-blown VM.  configure
> didn't like it :)
> 
> The problem is that configure tries to figure out how to get ps(1) to output
> all processes poorly.  There are a few issues here.
> [...]
>      it will just present more metrics), this check seems very misguided.
> 
> In order to proceed further, I made a quick hack to configure.ac. 

I think its time to rework this part of the configure script - its becoming
increasingly a source of build failures and frustration, and continuing to
add code to it may not be the long-term-healthy thing to do.

The main reason this configure.ac code exists is to figure out the arguments
needed for a platforms ps(1) to give arguments (i.e. set $pcp_ps_all_flags)
so that PCP scripts that need to run ps can do so in a platform-independent
way.

We currently attempt to do this dynamically, by running ps and sniffing the
output.  There is a section at the start of the configure script which has
platform-specific variables which are not dynamic.  I suggest we find the
"correct" values for pcp_ps_all_flags for each platform we can build on and
add the setting there, explicitly, and remove the dynamic check entirely.

Ken, AFAICT, nothing uses $pcp_ps_have_bsd - do you remember who/what that
was for?  Do we still need this, or can it be safely removed?

> Anyway, with this and the libzfs related change in my previous email pcp
> compiles and runs.  Now I'm off to figure out how to run pcpqa - I never
> actually ran it before and it looks a bit daunting.

qa/README is your friend.

cheers.

--
Nathan

<Prev in Thread] Current Thread [Next in Thread>