Buildfarm + Git tryouts

Started by Chris Browneover 15 years ago3 messages
#1Chris Browne
cbbrowne@acm.org

I'm trying to start preparing buildfarm nodes for the upcoming Git
migration, and have run into a few issues. I speculate that -hackers
is one of the better places for this to get discussed; if it should be
elsewhere, I'm sure Andrew Dunstan won't be shy to redirect this :-).

What I was hoping to do was to run nodes for a little while against
the repo at git://github.com/oicu/pg-cvs-mirror.git to validate that
it all works against git.

The best instructions available thus far...
<http://wiki.postgresql.org/wiki/PostgreSQL_Buildfarm_Howto&gt;

Unfortunately, I'm finding some anomalies in terms of differences
between my environments and what that suggests.

1. git clone --mirror is apparently new in git version 1.6

<http://kerneltrap.org/mailarchive/git/2008/8/2/2793244&gt;

My somewhat mouldy Ubuntu node is on git 1.5.4, which doesn't
include this.

There's a workaround, as --mirror is just "syntactic sugar"

$ git clone --mirror $URL
may be replaced by
$ git clone --bare $URL
$ (cd $(basename $URL) && git remote add --mirror origin $URL)

I see options:
a) Require git 1.6 or so (latest stable is 1.7.1.1)
b) Offer the alternative mirror approach

2. SCM.pl seems to expect the repository directory to be called
"pgsql" (around line #373)

I wound up hacking up the code to head to a specific directory,
which isn't a terribly good hack, but apparently works for now.

-               chdir 'pgsql';
+               chdir '/opt/build-farm/pgsql.git';

It feels as though something's not quite right about how the
nearby Git repository is referenced.

3. Some problems checking status.

i) Status Line: 491 bad ts parameter - [timestamp omitted] is in the future

I know my clock's reasonable - ntp is reporting I'm within 0.25s of
some stratum 2 nodes. Is it possible that the buildfarm server is
ill-synced?

ii) Status Line: 460 script version too low

I just pulled a buildfarm build last week, so I'm not sure what this
would relate to.

That's where I've gotten thus far. Node caracara isn't reporting in
just now because of this testing. I hope other people ought to be
doing similar tryouts so that the revision effort to shift to Git
doesn't wait to *start* until the development repo shifts.
--
let name="cbbrowne" and tld="gmail.com" in String.concat "@" [name;tld];;
http://linuxfinances.info/info/linuxdistributions.html
Rules of the Evil Overlord #77. "If I have a fit of temporary insanity
and decide to give the hero the chance to reject a job as my trusted
lieutentant, I will retain enough sanity to wait until my current
trusted lieutenant is out of earshot before making the offer."
<http://www.eviloverlord.com/&gt;

#2Alvaro Herrera
alvherre@commandprompt.com
In reply to: Chris Browne (#1)
Re: Buildfarm + Git tryouts

Excerpts from Chris Browne's message of lun jul 05 12:33:49 -0400 2010:

3. Some problems checking status.

i) Status Line: 491 bad ts parameter - [timestamp omitted] is in the future

I know my clock's reasonable - ntp is reporting I'm within 0.25s of
some stratum 2 nodes. Is it possible that the buildfarm server is
ill-synced?

I asked around and was told that the server is in sync currently. It
has been known to be out of sync before; if this problem still comes up,
maybe there's a software bug or something.

#3Andrew Dunstan
andrew@dunslane.net
In reply to: Alvaro Herrera (#2)
Re: Buildfarm + Git tryouts

Alvaro Herrera wrote:

Excerpts from Chris Browne's message of lun jul 05 12:33:49 -0400 2010:

3. Some problems checking status.

i) Status Line: 491 bad ts parameter - [timestamp omitted] is in the future

I know my clock's reasonable - ntp is reporting I'm within 0.25s of
some stratum 2 nodes. Is it possible that the buildfarm server is
ill-synced?

I asked around and was told that the server is in sync currently. It
has been known to be out of sync before; if this problem still comes up,
maybe there's a software bug or something.

This discussion really belongs on pgbuildfarm-members, I think.

I have added a small fudge factor to the time test - let's see if that
changes things.

cheers

andrew