pcp
[Top] [All Lists]

Re: [pcp] Build/install status for 3.7.0

To: Nathan Scott <nathans@xxxxxxxxxx>
Subject: Re: [pcp] Build/install status for 3.7.0
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Thu, 07 Mar 2013 07:00:09 +1100
Cc: pcp@xxxxxxxxxxx
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <1745058060.15558832.1362561319979.JavaMail.root@xxxxxxxxxx>
References: <1745058060.15558832.1362561319979.JavaMail.root@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130221 Thunderbird/17.0.3
On 06/03/13 20:15, Nathan Scott wrote:
On comma 3.6.11 i386 Darwin 10.8.0, the build of the logger PMDA
fails

Fixed now (dev 76becf8).

Confirmed, thanks.

On vm18 3.6.11 x86_64 LinuxMint 12 (lisa) the build dies a horrible
death in the python code with ...

I *think* this is from pyversions -nr returning more versions of
python than are installed on the machine.  I think in turn this
is because the build dependency is not ensuring python2.6 is there.
I've attempted a fix: dev b8f72f3 not 100% certain though - seems
like it might work.

OK this is more broken than I thought ... pyversions -vr (no -n) returns as follows:

LinuxMint _sometimes_
2.6 2.7
but most of time
pyversions: error parsing Python-Version attribute

On all the other Debian and Ubuntu systems the answer is
pyversions: error parsing Python-Version attribute

_if_ it returns "2.6 2.7" the build breaks, otherwise the for loop in debian/rules is not run (which also seems bad).

This is not doing anything to make we want to learn python!

...
lock.c:94:12: warning: ignoring return value of âstrerror_râ,
declared with attribute warn_unused_result [-Wunused-result]


Need to be careful here - strerror_r on Mac OS X returns int,
but char* on Linux.  These could be replaced with pmErrStr()
calls perhaps?

Yep, I'll change these to use pmErrStr_r() which will work just fine in this context.

On vm04 3.7.0 i586 CentOS 5.9 (Final)
>> ...
the linux pmda
pmda.c:732: warning: ârightnowâ may be used uninitialized in this
function


Fixed.  (dev 083a173)

Confirmed, thanks.

On vm05 3.7.0 i486 Gentoo 2.0.3 pmcd dies


Starting pmcd ...
[Wed Mar 6 16:13:51] pmcd(1293) Error: OpenRequestSocket(44321,
INADDR_ANY) __pmCreateSocket: Address family not supported by
protocol
[Wed Mar 6 16:13:51] pmcd(1293) Error: pmcd: can't open any request
ports, exiting
[Wed Mar 6 16:13:51] pmcd(1293) Error: pmcd not started due to
errors!

Hmm, this is probably the most concerning.  What does ifconfig on
this host report?  This looks like it might be an IPv6-related
regression, possibly, Dave?

Nothing obviously interesting from ifconfig ...

kenj@vm05:~$ /sbin/ifconfig -a
eth0      Link encap:Ethernet  HWaddr 52:54:00:73:7d:3a
          inet addr:192.168.1.205  Bcast:192.168.1.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:15213 errors:0 dropped:0 overruns:0 frame:0
          TX packets:480 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:2225766 (2.1 MiB)  TX bytes:78938 (77.0 KiB)
          Interrupt:11

lo        Link encap:Local Loopback
          inet addr:127.0.0.1  Mask:255.0.0.0
          UP LOOPBACK RUNNING  MTU:16436  Metric:1
          RX packets:12 errors:0 dropped:0 overruns:0 frame:0
          TX packets:12 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
          RX bytes:600 (600.0 B)  TX bytes:600 (600.0 B)

And detailed diags are not helping (hint, hint, ...)

kenj@vm05:~/src/pcp/src/pmcd/src$ ./pmcd -Dall -f
AddRequestPort: INADDR_ANY port 44321
__pmSetSocketIPC: fd=3
IPC table fd(PDU version):
pmGetConfig: PCP_TMP_DIR=/var/tmp
[Thu Mar 7 06:57:01] pmcd(1699) Error: OpenRequestSocket(44321, INADDR_ANY) __pmCreateSocket: Address family not supported by protocol


I'd like to get back to a point where we get some Windows servers
into the mix, somehow, but really don't have a good way to reach
that point (licensing issues, packaging issues, etc).  Perhaps we
could approach the Aconex folks for assistance with getting a
licensed VM into the farm?  (and then just keep git builds rolling
forward until an opportunity to tackle packaging next presents).

I have a Windows 8 VM there with a pre-release developer licence (that may or may not still work), but the biggest obstacle is there is no reliable recipe for creating a build environment with MSYS et al to create installable packages.

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