> On Fri, May 09, 2008 at 12:45:19AM -0400, ianb@... wrote:
> > Maybe poach-eggs --freeze would be useful to save information about the
> > build after it's been built?  It adds in those @rev things to svn urls.

Yep, this should be useful:

http://pw.openplans.org:6020/builders/opencore-minimal/builds/647/steps/shell_7/logs/stdio


-- 

Paul Winkler
http://www.openplans.org/people/slinkp/profile
yahoo: slinkp23
AIM:   slinkp1970

Paul Winkler wrote:
> Okay, so independently of getting the info from poach-eggs --freeze,
> would you object to me merging my fassembler branch?
> or do you think it's the just the wrong thing to do?
> https://svn.openplans.org/svn/fassembler/branches/slinkp-revision-option-20080508

i don't know what ian's thoughts are, but it feels a bit off to me.  maybe 
that's b/c i think of fassembler as more of a general tool, sort of like make 
or buildit.

of course, the fassembler source code is littered w/ TOPPisms... all of our 
custom projects are defined in there.  my druthers would be for the fassembler 
core to be its own project, with all of the topp:* projects defined in a 
separate package.

this is also part of my motivation for putting the code that manages our 
particular deployment specifics (directory structure, etc.) into a separate 
package, rather than rolling it into a part of fassembler's core.

of course, we may not have making fassembler work as a more general tool on 
our list of cares; it's certainly not a high organizational priority.  even 
then, though, i'd still lean against adding arbitrary variable options to the 
countless we already have, if you've got a solid workaround that makes it 
unnecessary.

2-cents-which-won't-get-you-much-these-days'ly yours,

-r


> 
> - PW
> 
> On Fri, May 09, 2008 at 12:45:19AM -0400, ianb@... wrote:
>> In a requirement file you can add @rev-number to any svn URL.  Would that
>> satisfy your need?
>>
>> Maybe poach-eggs --freeze would be useful to save information about the
>> build after it's been built?  It adds in those @rev things to svn urls.
>>
>>
>>
>> On Thu, May 8, 2008 2:27 pm, Paul Winkler wrote:
>>> Today I was wishing that I could force fassembler to use a specific
>>> revision number for things it checks out from svn.openplans.org.
>>>
>>> This is because when buildbot builds fail, it's often tricky to
>>> determine exactly what revision was built, since it takes several minutes
>>> between when the build starts and when various things get checked out.
>>> Often, more commits are made during that time, and these
>>> don't make it to the blamelist, and I have to guess what was actually built
>>> and tested. And by the time I hear about the failure, the environment has
>>> likely been wiped out for the next build.
>>>
>>> So I hacked up this branch:
>>> https://svn.openplans.org/svn/fassembler/branches/slinkp-revision-option-2
>>> 0080508
>>>
>>>
>>> ... which adds an option to fassembler:topp like so:
>>>
>>>
>>> ./bin/fassembler fassembler:topp openplans_svn_revision=15000 ...
>>>
>>>
>>> If specified, this is used for the requirements checkout (if it's on
>>> svn.openplans.org), and specifies the same revision for any editable
>>> checkouts in *-req.txt, regardless of any revision information in those
>>> (again, only for things on svn.openplans.org).
>>>
>>>
>>> This works AFAICT, within some IMO reasonable limitations:
>>>
>>>
>>> * other fassembler: projects don't recognize the option on the command
>>> line, so you'd have to edit build.ini if you change your mind later - or
>>> re-run fassembler:topp.
>>>
>>> * re-running fassembler:opencore with a changed revision in build.ini
>>> can fail if things have moved around in the repo; first updating
>>> requirements/ to that revision should help
>>>
>>> * things could fail at runtime if there's some incompatibility between
>>> the revision and an updated third-party component; hopefully this should be
>>> rare, unless maybe you're building really old revisions for some reason.
>>>
>>> Does anybody object to anything here? If not I'll merge it to the
>>> fassembler trunk, and then update buildbot to take advantage of it.

Paul Winkler wrote:
> Okay, so independently of getting the info from poach-eggs --freeze,
> would you object to me merging my fassembler branch?
> or do you think it's the just the wrong thing to do?
> https://svn.openplans.org/svn/fassembler/branches/slinkp-revision-option-20080508

I'd rather this take the form of something like 
requirements/revision-pin.txt, which would contain:

   https://svn.openplans.org/svn/ {YYYYMMDD}

(Any revision specifier that svn took would work, like an integer or a date)

I think this would be both more general and more useful.  Or, maybe 
instead of an ad hoc format in a file like this, something like 
requirements/control.ini:

   [repository https://svn.openplans.org/svn/]
   pin_revision = {YYYYMMDD}

Then we could add more options later.  E.g., if we make a local mirror, 
we could do:

   [repository http://svn.zope.org]
   mirror = https://svn.openplans.org/mirrors/svn.zope.org

Sometimes a more general option is desired, like:

   [repository *]
   pin_revision = {YYYYMMDD}

which would be a nice catch-all (jhbuild has an option like this, for 
example).

   Ian