You are viewing a plain text version of this content. The canonical link for it is here.
Posted to general@gump.apache.org by "Adam R. B. Jack" <aj...@trysybase.com> on 2003/10/07 17:38:40 UTC
Python Gump update
All,
First, thanks for the feedback on the "please read/reply" mail. The input
really helped to see through the eyes of other's & take a bit of a step
back. For one, I said "I don't want to walk the Gump path alone", and I
don't -- things benefit from the insights of a group, but I will if needs
be. Also, I'm not alone -- I'm just alone (today) pounding Python code, you
guys are contributing knowledge/insights/resources/graphics, those are all
invaluable.
I came to writing OSS because I use/deploy so much OSS that I benefit mostly
from other's development efforts, that I want to help them work together in
the real world, (see http://www.krysalis.org/version/jar-hell.html). I want
to support others in their efforts, support the community, support scalable
re-use -- or (at least) shift the balance a little that way. Uncool, or
infrastructure, or whatever -- Gump is a way to help re-use, and I believe
in that, I think it is cool. As such, I will continue with Python Gump as
long as I am making progress. I hope to take it so far that becomes the
defacto Gump due to merit.
With big thanks to Leo Simons, Scott Sanders, Nick Chalko, for support we
have a few running Python Gumps, two full and one partial:
LSD @ Midnight CST/CEST = http://lsd.student.utwente.nl/gump/index.html
DotNot @ 05:00PST/PDT = http://gump.dotnot.org/
Chalko = http://gump.chalko.com (Temporary network outage as of time
of writing)
New things:
1) I've added a command timeout (defaults to 30 minutes) that eventually
I'll allow to be project configurable. [This is *nix only, not M$].
[I am trying not to mess w/ the project descriptors yet, so leaving things
like that to later.]
2) I've split pages. We now have TODOs (Modules with issue) and Modules
(all) and Packages:
http://gump.dotnot.org
http://gump.dotnot.org/todos.html
http://gump.dotnot.org/modules.html
http://gump.dotnot.org/packages.html
[I still need to figure out Forrest site.xml & get a nice skin, but I'm
getting there. Nicola helped w/ some cool icons.]
3) I keep adding more and more to documentation (in the main to help me
debug).
Notice :
1) The annotations (log messages) at top of the project
2) The classpath. annotated with what contributed it (i.e from
dependencies).
http://gump.dotnot.org/krysalis-centipede/krysalis-centipede.html
4) I found that somebody had attempted to implement the various inherit
mechanisms in <depend.
Namely inherit = none/all/runtime/jars/hard. Unfortunately they gave up and
only implemented 'jars',
hence a bunch of builds were failing. I don't feel comfortable attempting to
implement all these choices right now (I'd like to do some clean-up first)
so I've attempted to switch it to 'all' for now. Crude, but hopefully
workable.
Anybody have any views on if all these options are still needed? I did a
troll through the projects, and some are hardly used. Just a question...
----------------------------------------------------------------------------
-------------
Observations:
1) SF.net CVS is still a major pain, and is causing much not to build.
[Junit being at SF.net many tests fail.]
2) I may still have CVS password bugs. I tried copying the new password file
format, I'll debug more.
3) I think I need to pick one "failed project" a day, compare it to a
traditional gump run, and if that worked, then investigate/debug. This will
take time, but there is so much output from a gump that I need such an
approach. Please be patient, I'll get through the issues.
----------------------------------------------------------------------------
-------------
I still have features to implement (e.g. gather the outputs) but I think I'm
past a lot of the 'processing logic' problems, and that this Gump is
basically working. I'd appreciate feedback on the interface, if anybody has
time to walk a site. What things are missing, what are hard to navigate,
etc.
I'll probably enter some of this status into the Jakarta Gump entry in the
newsletter, so long as nobody has any objections.
regards
Adam
--
Experience Sybase Technology...
http://www.try.sybase.com
SF.net CVS -- one more week, we hope.
Posted by "Adam R. B. Jack" <aj...@trysybase.com>.
Leo wrote:
> maybe we should depend on binary versions of some of the vital packages
> (like JUnit) hosted at SF until they finish upgrading their hardware
> (they're installing lots).
>
On SF.net I see this below. Let's give them another week or so.
regards
Adam
http://sourceforge.net/docman/display_doc.php?docid=2352&group_id=1#cvs
(2003-10-10 09:22:56) Roughly 40% of project CVS repositories are now
online using our new CVS infrastructure; we expect this transition to be
completed during the week of 2003-10-13. Additional details, including
details of the benefits of the new infrastructure, will be posted to Site
Status pending completion of this migration.
Re: Python Gump update
Posted by "Adam R. B. Jack" <aj...@trysybase.com>.
> that is just way cool! I just spent some time browsing the latest
> generated stuff. It sure is eye-pleasing :D
Thanks. Forrest/forresters get that credit. :-) I'm glad though, 'cos I
think that there are times when a pleasing presentation does open minds to
the presenter. I've tried to put as much information out as possible, and
will (over time) whittle it down to the most important first.
> maybe we should depend on binary versions of some of the vital packages
That is a good idea, especially since JUnit is (1) pretty darn stable (2)
typically runtime.
> (like JUnit) hosted at SF until they finish upgrading their hardware
> (they're installing lots).
<ranton>
Yes, but are the making progress, or just a miserable mess? This has gone on
for months. They are way past their revised completion date & I don't see
much good happening. I sent a 'can you give us update please' e-mail to the
gent whose name was one (he purposly published his e-mail account asking for
such) and all I got was an autoreply that he was out most of this month.
Gak! :(
</rantoff>
> I love the icons and the fact that absolutely every data is available in
> browsable html. What is not so simple is coming into the front page and
> quickly seeing what the major issues are at the particular moment. With
> the 'old gump', it takes 2 page down hits and a single click to figure
> out what went wrong to any project that interests me.
Completely agreed. Right now I spew everything for debugging (the Python
part) but I do think it'll help for for debugging their builds. I'll work on
a easier navigation & presentation shortly. Thanks for the feedback.
> A way to improve on that might be to make the "X" icon link to the build
> output, iow make the build failed icon for xalan link to
>
>
http://lsd.student.utwente.nl/gump/xml-xalan/build/build_xml-xalan_xml-xalan2.html#Output
Love the idea. Nick keeps telling me this also, and I laugh 'cos I know I
keep trying to click on those icons also. I'll get there, it is quite
doable.
> I can imagine it must be even more painful navigating the site for
> people with low bandwidth connection (I mean, my connection is over
> 100mbit ethernet), especially since the forrest skin is a bit on the
> heavy side.
Funny, the skin doesn't seem to bother me (Mr Modem user). The top comes
down very fast. I'm not sure the skin is as 'heavy' as you think. That, or
I've just gotten used to mollasses...
regards
Adam
Re: Python Gump update
Posted by Leo Simons <le...@apache.org>.
> http://gump.dotnot.org/todos.html
that is just way cool! I just spent some time browsing the latest
generated stuff. It sure is eye-pleasing :D
> 1) SF.net CVS is still a major pain, and is causing much not to build.
> [Junit being at SF.net many tests fail.]
maybe we should depend on binary versions of some of the vital packages
(like JUnit) hosted at SF until they finish upgrading their hardware
(they're installing lots).
> I'd appreciate feedback on the interface, if anybody has
> time to walk a site. What things are missing, what are hard to navigate,
> etc.
I love the icons and the fact that absolutely every data is available in
browsable html. What is not so simple is coming into the front page and
quickly seeing what the major issues are at the particular moment. With
the 'old gump', it takes 2 page down hits and a single click to figure
out what went wrong to any project that interests me.
A way to improve on that might be to make the "X" icon link to the build
output, iow make the build failed icon for xalan link to
http://lsd.student.utwente.nl/gump/xml-xalan/build/build_xml-xalan_xml-xalan2.html#Output
I can imagine it must be even more painful navigating the site for
people with low bandwidth connection (I mean, my connection is over
100mbit ethernet), especially since the forrest skin is a bit on the
heavy side.
> I'll probably enter some of this status into the Jakarta Gump entry in the
> newsletter, so long as nobody has any objections.
Go for it!
- LSD
Re: Charting ( was Re: Python Gump update)
Posted by Nicola Ken Barozzi <ni...@apache.org>.
Adam R. B. Jack wrote:
>>>BTW, could you add some description to the FOG Factor statistics (what
>>>is it, how is it calculated, is -10 better than 10 ...)
>>
>>In addition, it would be swell if you start already logging them
>>somewhere. I can add a chart-making part to the system.
>
> Please tell me more. I have these things in a DBM file database, so I can
> output anything you need. What formats can you take and chart?
Look at http://jcharts.krysalis.org/ for example of charts.
If you can spit out in the future the chart data as XML, I can make
forrest pick that up and make a chart. Any xml format will do for now,
just to see if your output can be ok.
--
Nicola Ken Barozzi nicolaken@apache.org
- verba volant, scripta manent -
(discussions get forgotten, just code remains)
---------------------------------------------------------------------
Charting ( was Re: Python Gump update)
Posted by "Adam R. B. Jack" <aj...@trysybase.com>.
> > BTW, could you add some description to the FOG Factor statistics (what
> > is it, how is it calculated, is -10 better than 10 ...)
>
> In addition, it would be swell if you start already logging them
> somewhere. I can add a chart-making part to the system.
Please tell me more. I have these things in a DBM file database, so I can
output anything you need. What formats can you take and chart?
regards
Adam
Re: Python Gump update
Posted by Nicola Ken Barozzi <ni...@apache.org>.
Stefan Bodewig wrote:
> On Tue, 7 Oct 2003, Adam R. B. Jack <aj...@trysybase.com> wrote:
...
>>I'll probably enter some of this status into the Jakarta Gump entry
>>in the newsletter, so long as nobody has any objections.
>
> +1
>
> BTW, could you add some description to the FOG Factor statistics (what
> is it, how is it calculated, is -10 better than 10 ...)
In addition, it would be swell if you start already logging them
somewhere. I can add a chart-making part to the system.
--
Nicola Ken Barozzi nicolaken@apache.org
- verba volant, scripta manent -
(discussions get forgotten, just code remains)
---------------------------------------------------------------------
FOG Factor (was Re: Python Gump update )
Posted by "Adam R. B. Jack" <aj...@trysybase.com>.
> BTW, could you add some description to the FOG Factor statistics (what
> is it, how is it calculated, is -10 better than 10 ...)
I used the one I proposed on this list eons ago, the simple "successful
builds - (fails + prereq fails)". The higher (non-negative) the number the
better. I'm more than open to modify this in the future, it is just a quick
stat to eyeball. I'll add this information to the page.
Note: Right now it is meaningless 'cos it is more a testament to Gumpy
coding than to the project. I'll reset this shortly.
regards
Adam
Re: Python Gump update
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 7 Oct 2003, Adam R. B. Jack <aj...@trysybase.com> wrote:
> 4) I found that somebody had attempted to implement the various
> inherit mechanisms in <depend. Namely inherit =
> none/all/runtime/jars/hard.
>
> Anybody have any views on if all these options are still needed?
All of them AFAICT.
> I'll probably enter some of this status into the Jakarta Gump entry
> in the newsletter, so long as nobody has any objections.
+1
BTW, could you add some description to the FOG Factor statistics (what
is it, how is it calculated, is -10 better than 10 ...)
Cheers
Stefan
Re: Python Gump update
Posted by Scott Sanders <sc...@dotnot.org>.
> I'll probably enter some of this status into the Jakarta Gump entry in
> the
> newsletter, so long as nobody has any objections.
>
>
Absolutely.
Scott