You are viewing a plain text version of this content. The canonical link for it is here.
Posted to general@gump.apache.org by Bill Barker <wb...@wilshire.com> on 2005/06/07 06:52:05 UTC
Gump runs stopped?
According to http://vmgump.apache.org/gump/public/index.html, Gump hasn't
run since Midnite last Friday. Is is down for a reason?
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 07 Jun 2005, Leo Simons <ma...@leosimons.com> wrote:
> Any chance you have time to get your classic gump installed over
> onto the solaris gump zone along
Once I manage to log in, yes. Sorry, I didn't try before (see your
mailbox for details).
> with some instructions on how to run that check?
Sure. I do have the latest versions of the XSLT stylesheets and all
that around, so it would be a matter of running
$ bash gen.sh
> We could have a cronjob to update from CVS, run the check, and send
> e-mail if it finds a problem.
Should be simple to extract something like that from gen.sh's merge
step.
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by "Adam R. B. Jack" <aj...@apache.org>.
Gak, damn. Bad to worse. Working on it...
Adam
----- Original Message -----
From: "Stefan Bodewig" <bo...@apache.org>
To: <ge...@gump.apache.org>
Sent: Tuesday, June 07, 2005 2:57 PM
Subject: Re: Gump runs stopped?
> On Tue, 7 Jun 2005, Adam R. B. Jack <aj...@apache.org> wrote:
>
> > In short, no quick fix comes to me.
>
> Could you please revert your last commit? It has created tons of
> false nag mails because Gump now thinks there'd be circular
> dependencies where none are there.
>
> Stefan
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
> For additional commands, e-mail: general-help@gump.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 7 Jun 2005, Adam R. B. Jack <aj...@apache.org> wrote:
> In short, no quick fix comes to me.
Could you please revert your last commit? It has created tons of
false nag mails because Gump now thinks there'd be circular
dependencies where none are there.
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by Leo Simons <ma...@leosimons.com>.
On 07-06-2005 22:09, "Adam R. B. Jack" <aj...@apache.org> wrote:
>> . Adam, could you give
>> some pointers (filenames and linenumbers) where you think the insertion
>> point for a cyclic dependency checker should be in gump3?
>
> I think Gump3 has some such code in there already.
Dammit. I meant Gump2. Sorry :-$
> Trouble is, that fails to publish this information to the masses.
Right. Which is why I thought we might plug that code into Gump *2*, *now*.
Sorry again.
LSD
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by "Adam R. B. Jack" <aj...@apache.org>.
>. Adam, could you give
> some pointers (filenames and linenumbers) where you think the insertion
> point for a cyclic dependency checker should be in gump3?
I think Gump3 has some such code in there already.
What I think is missing is an (equally supported) use case of "check
metadata" that is as well used/supported as "build" or "update and build".
Basically we need a code path (and/or algorithm/set of plugins) that
verifies the metadata & (perhaps) notifies and/or 'documents' (to DB?).
The main reason I added --do-builds (to match --do-updates) -- so when
neither is specified it does a verify. I was to use what is there for this
purpose. Trouble is, that fails to publish this information to the masses.
It'll do for now, but it needs a re-think.
Gump2 fell down here 'cos it tried to objectify all the metadata, and then
annotate it with errors (e.g. circular dependency here.) This meant the
object tree was "untrustworthy", containing bogus references/object, hence
the crashes and hacks. Gump3's approach to drop bad things makes the tree
trustworthy, but lacking in this bad information, so Gump3 can't use it's
plug-ins to publish.
In short, no quick fix comes to me.
regards,
Adam
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by Leo Simons <ma...@leosimons.com>.
On 07-06-2005 21:16, "Stefan Bodewig" <bo...@apache.org> wrote:
> On Tue, 07 Jun 2005, Stefan Bodewig <bo...@apache.org> wrote:
>> On Tue, 7 Jun 2005, Adam R. B. Jack <aj...@apache.org> wrote:
>
>>> Ok, the what-is-becoming usual, i.e. circular dependency in the
>>> metadata (dom4j/jaxen, I believe).
>>
>> "classic Gump" doesn't flag one.
>
> Bzzt wrong. Don't use a stalled source tree.
>
> Fixed now.
Stefan,
Any chance you have time to get your classic gump installed over onto the
solaris gump zone along with some instructions on how to run that check? We
could have a cronjob to update from CVS, run the check, and send e-mail if
it finds a problem.
---
I could probably have a look at getting the gump walker+verifier modified to
walk the gump2 code. The algorithm is pretty generic. Adam, could you give
some pointers (filenames and linenumbers) where you think the insertion
point for a cyclic dependency checker should be in gump3?
Cheers!
Leo
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 07 Jun 2005, Stefan Bodewig <bo...@apache.org> wrote:
> On Tue, 7 Jun 2005, Adam R. B. Jack <aj...@apache.org> wrote:
>> Ok, the what-is-becoming usual, i.e. circular dependency in the
>> metadata (dom4j/jaxen, I believe).
>
> "classic Gump" doesn't flag one.
Bzzt wrong. Don't use a stalled source tree.
Fixed now.
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 7 Jun 2005, Adam R. B. Jack <aj...@apache.org> wrote:
>> Me too. I can't think of a good reason why it ought hang, and I
>> couldn't find a cause when I last saw a hang (just before). I fear
>> this might be happening every time.
>
> Ok, the what-is-becoming usual, i.e. circular dependency in the
> metadata (dom4j/jaxen, I believe).
"classic Gump" doesn't flag one.
Could it be the ext-CVS over SSH for icu4j? This doesn't work anyway,
we should replace icu4j with an installed package.
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by "Adam R. B. Jack" <aj...@apache.org>.
> Me too. I can't think of a good reason why it ought hang, and I couldn't
> find a cause when I last saw a hang (just before). I fear this might be
> happening every time.
Ok, the what-is-becoming usual, i.e. circular dependency in the metadata
(dom4j/jaxen, I believe). Clearly given the complexities of project
interactions (and circular dependencies in the real OSS world, not just Gump
metadata) this is something Gump needs to handle a lot nicer. I've promoted
my checks for that (and tested/fixed them for this case) and hope we stay a
little healthier w.r.t metadata errors.
Gump3 is a little cleaner w.r.t these issues, but since it is nowhere near
ready for prime time, that is somewhat moot. That, and we really need a
simply tool for validating metadata, so we don't have to debug/guess when
Gump runs fail.
Anyway, there is a new run going, hopefully it'll keep going this week (as I
head off for a long weekend camping at the Great Sand Dunes, Colorado's
"beach".)
regards,
Adam
P.S. This (clearly) was nothing to do with the VmGump move, it is just
coincidental timing.
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by "Adam R. B. Jack" <aj...@apache.org>.
> On Mon, 6 Jun 2005, Bill Barker <wb...@wilshire.com> wrote:
>
> > According to http://vmgump.apache.org/gump/public/index.html, Gump
> > hasn't run since Midnite last Friday. Is is down for a reason?
>
> There was a hanging Gump instance from June 5th, I've killed that and
> asserted that no lock files have been kept.
>
> I hope it is going to come back to live soon.
Me too. I can't think of a good reason why it ought hang, and I couldn't
find a cause when I last saw a hang (just before). I fear this might be
happening every time.
Amazing how the move from Brutus to VmGump has been so 'uncomfortable'.
regards
Adam
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Gump runs stopped?
Posted by Stefan Bodewig <bo...@apache.org>.
On Mon, 6 Jun 2005, Bill Barker <wb...@wilshire.com> wrote:
> According to http://vmgump.apache.org/gump/public/index.html, Gump
> hasn't run since Midnite last Friday. Is is down for a reason?
There was a hanging Gump instance from June 5th, I've killed that and
asserted that no lock files have been kept.
I hope it is going to come back to live soon.
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org