buildbot / buildbot Goto Github PK

Python-based continuous integration testing framework; your pull requests are more than welcome!

License: GNU General Public License v2.0

Makefile 0.15% Python 88.90% Shell 0.12% HTML 0.06% JavaScript 0.15% RAML 0.91% Roff 0.13% TypeScript 8.98% Dockerfile 0.02% Pug 0.01% Less 0.01% Jinja 0.06% Mako 0.01% CSS 0.01% SCSS 0.46% Gherkin 0.03%

continuous-integration ci python ci-framework

buildbot's Introduction

Buildbot

The Continuous Integration Framework

Buildbot is based on original work from Brian Warner <mailto:warner-buildbot @ lothar . com>, and currently maintained by the Botherders.

Visit us on http://buildbot.net !

_ _ _

Buildbot consists of several components:

master
worker
www/base
www/console_view
www/waterfall_view

and so on

See the README in each subdirectory for more information

Related repositories:

https://github.com/buildbot/buildbot-media - Buildbot-related media
https://github.com/buildbot/buildbot-website - Source for http://buildbot.net

buildbot's People

Contributors

Stargazers

Watchers

Forkers

djmitche dustin garyposter jgraff deejay1 gward gescalante mahatmamanic lambda codeyman2 warner elemel douglatornell astraw knielsen matthewryanscott mue-zz gigatropolis adamac azverkan tangerilli nsylvain aloisiojr dhubler lurker99 mdirolf ktf rajveerr mgedmin tfogal verm abyx benallard pythonologist garetharmstronghp satyagraha thatch gvalkov paperclip chasephillips giallu frelo oppianmatt jollyroger leatherslave1 oberstet jaysoffian marcmengel tvo pike almad perilousapricot nicolas17 liucougar eklitzke ssalbiz djs pao thomascantrell unpaidbill timothyfitz andyhowell martinez wener abompard lool shawnchin tardyp ayust dwlocks thmo clepple kami loewis strk cread tomprince dionorgua jeisinger usualoma philippem dbs jmartens geokollias lazerdye jamesob kaladron dgladkov ahussein jamesnkh sillern lsblakk smackware hef stsquad sagardh littleforker thomasvs jfschedler paroga

buildbot's Issues

gerritStartCB don't work

This ticket is a migrated Trac ticket 2755

People contributed to the original ticket: @djmitche, @unknown_contributor
Ticket created on: Apr 13 2014
Ticket last modified on: Jun 18 2014

Hi,

I try the code in http://docs.buildbot.net/latest/manual/cfg-statustargets.html about update Gerrit Status.

No problem with gerritReviewCB or gerritSummaryCB with last trunk
but gerritStartCB don't work

I look the code
and in master/buildbot/status/status_gerrit.py

        # Gerrit + Git
        if build.getProperty("gerrit_branch") is not None:  # used only to verify Gerrit source

build.getProperty("gerrit_branch") is always None with [[GerritStart]]CB

Git(mode="copy", ..) doesn't work with multiple codebases

This ticket is a migrated Trac ticket 2821

People contributed to the original ticket: @djmitche, @unknown_contributor
Ticket created on: Jun 11 2014
Ticket last modified on: Jun 18 2014

Hello,

I am using multiple git repositories, some of which derive data from another. For instance, for a release, the changelog of the software is split into posts for the blog.

Another use case is to make performance measurements, and to commit the result in a performance web site git.

I would like to have:

Git(repo=repo1, workdir # w2, mode "copy")
Git(repo=repo2, workdir # w2, mode "copy")

and then the ability to make
[[ShellCommand]](workdir # w2, command "xxx %s" % w1" )

so that build step scripts can work across repository boundaries.

Currently, I am working these limitations around by using git submodules, which scripts will automatically update, since git has no means to track branches of other repos as submodules.

To me, it would be cleaner, if these cross-repo scripts and builders were possible. Right now, I don't have an urgent need, as I believe submodules are workable, with the noise that the automatic submodule updates bring, it's not an optimal solution, and bears no flexibility. I cannot e.g. override from the outside the commit id values for these submodules.

Comment from: @dustin
Date: Jun 11 2014

Buildbot's source steps are intended for checkouts, not for making new commits.

That said, you can use multiple Git steps, with codebases allowing you to have those steps use different sourcestamps.

The ticket title says "Cannot", but doesn't give any evidence of why you cannot. Please re-open if that's not working for you for some reason.

Comment from: @dustin
Date: Jun 12 2014

Ah! I just read the corresponding mailing-list thread (http://article.gmane.org/gmane.comp.python.buildbot.devel/10311)

The issue here is that Git's "copy" mode is not codebase-compatible, per http://article.gmane.org/gmane.comp.python.buildbot.devel/10314.

Comment from: @dustin
Date: Jun 12 2014

Apologies for my confusion.

buildbot.test.unit.test_scripts_start.TestStart.test_start is flaky

This ticket is a migrated Trac ticket 2760

People contributed to the original ticket: @djmitche
Ticket created on: Apr 20 2014
Ticket last modified on: Aug 03 2014

2014-04-17 18:41:54-0400 [-] --> buildbot.test.unit.test_scripts_start.[[TestStart]].test_start <--
2014-04-17 18:42:04-0400 [-] Main loop terminated.

and

[FAIL]
Traceback (most recent call last):
  File "/var/lib/buildslave/bbot/py27-tw1110/build/master/buildbot/test/unit/test_scripts_start.py", line 109, in cb
    self.assertSubstring('[[BuildMaster]] is running', out)
  File "/var/lib/buildslave/bbot/py27-tw1110/sandbox-python2.7-Twisted==11.1.0samigr-0.7.1/lib/python2.7/site-packages/twisted/trial/unittest.py", line 421, in fail[[UnlessSubstring]]
    return self.fail[[UnlessIn]](substring, astring, msg)
twisted.trial.unittest.[[FailTest]]: '[[BuildMaster]] is running' not in "Following twistd.log until startup finished..\n\nThe buildmaster took more than 10 seconds to start, so we were unable to\nconfirm that it started correctly. Please 'tail twistd.log' and look for a\nline that says '[[BuildMaster]] is Running' to verify correct startup.\n\n"

The test is disabled in #2748.

Make GitHub hook listen to other kind of events as well

This ticket is a migrated Trac ticket 2887

People contributed to the original ticket: @djmitche, @sa2ajj
Ticket created on: Sep 07 2014
Ticket last modified on: Jun 22 2015

That would allow to build around Buildbot systems aware of a particular workflow.

Cf.: [[Bors] and [https://github.com/barosl/homu Homu|https://github.com/graydon/bors]]

Comment from: @dustin
Date: Sep 07 2014

This would be a powerful improvement, if the events and reactions to them were easily configurable.

GitHubStatus and HTTP Proxy Support?

Hi,

Is there an known way to make use of the GitHubStatus module if my server is sitting behind an HTTP proxy? I've also opened a ticket with the txgithub project asking the same question. So far I haven't figured out how to set up the twisted networking used by txgithub to filter through my proxy. Any help is appreciated.

Thanks

Support a single config across the cluster, or something like that

This ticket is a migrated Trac ticket 2959

People contributed to the original ticket: @djmitche, @sa2ajj, benoit.allard@...
Ticket created on: Oct 22 2014
Ticket last modified on: Feb 08 2015

The www interface is showing all builders that ever where there. This is a difference with eight were only the 'active' builders were displayed on the web interface. I believe the latter is wished.

Comment from: @dustin
Date: Oct 23 2014

This is a little tricky -- we don't want builders to "disappear" if the master that happens to be hosting them goes down. Although we could do that now, with a rather complex set of queries.

I suspect that the better plan will be to have builders disappear using the same kind of "horizon" cleanup we do for logs and builds.

Comment from: @ben
Date: Oct 23 2014

Provided this 'horizon' cleanup would be implemented in 0.9.+, this would then, if I understand correctly, prevent me to look at builders where build didn't happen for some time, even if the builders are still 'active' (defined in my config).

Anyway, the trouble I'm having now, is that at each reconfig where I renamed a builder, the old one stay there. Those were previous tests from me, and have no value anymore once the renamed builder is there.

I need a way to cleanup my interface, and anyone who experiment a bit with their configuration also does.

Comment from: @dustin
Date: Oct 24 2014

I wonder what would be the best interface for that. The options I see are:

horizons: delete builders which have no active master and no builds
immediate disappearance: if there are no active masters with a builder configured, it disappears
command-line cleanup: buildbot delete-builder XXX
web UI: administrators can delete builders from the web UI

I'm not sure which of these options is best..

Comment from: @sa2ajj
Date: Oct 24 2014

FWIW, two last options look the most appealing to me (cli & web ui options):

people do make mistakes -> give them a chance to see that something is wrong
in a multi-master setup it's more likely to make a mistake

Comment from: @ben
Date: Oct 25 2014

During a reconfig, we should be able to flag the dead builders as dead, or not ?

Comment from: @ben
Date: Oct 25 2014

Same is valid for the slaves by the way. Even worst, the slaves still show on the web interface a connection with the builders they were connected to (when they existed). I.e.: configured_on: [{"builderid":1,"masterid":1},{"builderid":2,"masterid":1},{"builderid":3,"masterid":1}] But that slave is absolutely not configured there ...

Those slave got new names / passwords, hence disapeared, and have been replaced by other ones ...

Comment from: @dustin
Date: Oct 26 2014

Reconfig happens on one master at a time, so it's still tricky at that point. Probably the worst outcome would be to have a builder deleted while its builderid was still cached in RAM in another master -- any new inserts with that ID would fail.

So, yes, I think a way to delete builders (and slaves) manually would be good.

Comment from: @ben
Date: Nov 07 2014

I believe, we need to move to a next-generation multi-master configuration.

The more we move stuff in the db, the more we are having this kind of troubles which boils down to:

We do have a decentralized system without coordinating entity (well, the db, but it's only passive).

What if we would require for the master.cfg to always contains the whole definition, and then, during buildbot start we would add a parameter to tell the master which one he actually is ?

Comment from: @dustin
Date: Nov 07 2014

What are some good models for this, among other clustered applications?

I think your suggestion is to use the same master.cfg for all masters, with runtime flags that cause it to be interpreted differently. But, how would you enforce that? And how would you accomplish a rolling configuration update?

LatentSlaveBuilder status does not get reset to LATENT when detached

This ticket is a migrated Trac ticket 2770

People contributed to the original ticket: @djmitche, @unknown_contributor, @sa2ajj
Ticket created on: Apr 30 2014
Ticket last modified on: Sep 04 2014

When insubstantiating a latent slave, its slavebuilders' status doesn't get reset to LATENT. On subsequent builds, the SlaveBuilder list passed to nextSlave will incorrectly note the slave as IDLE instead of LATENT.

Comment from: @sa2ajj
Date: Sep 04 2014

I am not sure I understand the desired behaviour: when the build is finished, but the buildslave is not shut down, it'd show IDLE. As soon as it's down, it'd become LATENT.

Comment from: @unknown_contributor
Date: Sep 04 2014

Clarifying the description a bit more:

In the list of SlaveBuilder objects passed to !nextSlave, the slave's status after finishing a build is IDLE. When a slave is insubstantiated, however, the slave's status remains as IDLE. The status attribute in the SlaveBuilder object is not reset to LATENT, the state it has before first substantiation.

Improve support for Gerrit

This ticket is a migrated Trac ticket 2890

People contributed to the original ticket: @sa2ajj
Ticket created on: Sep 09 2014
Ticket last modified on: Sep 09 2014

There's a number of things that could be improved for Gerrit support.

(And I'd need to create a ticket for this particular problem: force the version (if an installation is prepared incorrectly, Gerrit fails to report its version: might fail certain actions).)

List of relevant tickets

[[TicketQuery(status=new&status=assigned&status=reopened&keywords=~gerrit,group=type)]]
#2851
(missing other tickets)

MailNotifier 'all' mode does not work

When I set mode=['all'] on MailNotifier, I get an error in the log file:

2014-03-24 10:50:46-0400 [-] Configuration Errors:
2014-03-24 10:50:46-0400 [-]   mode all is not a valid mode

Failed FileUpload step generates a 0-sized file in the destination.

This ticket is a migrated Trac ticket 2826

People contributed to the original ticket: @djmitche, @unknown_contributor
Ticket created on: Jun 18 2014
Ticket last modified on: Jun 20 2014

I have a build step like this:

factory.addStep([[FileUpload]](slavesrc="file.png", workdir=dir_on_the_source, mode=0777, name="Upload", masterdest=[WithProperties], url=[WithProperties], halt[[OnFailure]]=False, flunk[[OnFailure]]=False, hide[[StepIf]]=lambda results, s: results == FAILURE))

Observed:
If the source file 'file.png' does not exist on the build slave, the build master will still create a new file to the destination, but the size of that file will be zero.

Also, if I remove the flunk[[OnFailure]], halt[[OnFailure]] and hide[[StepIf]] directives, the log will not indicate a reason for the failure (e.g. "File upload failed, because source file was not found")

Expected:

No zero-sized files should be generated in the destination and a descriptive error message should be displayed in the output.

status_json.py generates broken example links with path_to_root

If a master is running with root endpoint at http://www.foo.com/buildbot/ all the links on the json help pages that link to examples are broken.

move github_buildbot out of contrib

This ticket is a migrated Trac ticket 2828

People contributed to the original ticket: @djmitche, @sa2ajj, @ewongbb
Ticket created on: Jun 22 2014
Ticket last modified on: Jan 14 2017

This really should be a part of core Buildbot, as a web hook..

Comment from: @dustin
Date: Jun 22 2014

I especially like the addition of support for --secret, which allows Buildbot to at least trust that the request is from GitHub.

Comment from: @sa2ajj
Date: Sep 08 2014

buildbot.test.integration.test_slave_comm.TestSlaveComm.test_duplicate_slave is flaky

This ticket is a migrated Trac ticket 2761

People contributed to the original ticket: @djmitche, @tardyp
Ticket created on: Apr 20 2014
Ticket last modified on: Dec 14 2016

Traceback (most recent call last):
Failure: twisted.trial.util.[[DirtyReactorAggregateError]]: Reactor was unclean.
Selectables:
<Broker #1 on 28977>

Disabled in #2748.

Comment from: @tardyp
Date: Dec 14 2016

#2545

ln: failed to create symbolic link `/sbin/initctl': File exists

Step 2 : run ln -s /bin/true /sbin/initctl
---> Running in 6dbe7d164415
ln: failed to create symbolic link `/sbin/initctl': File exists

All work if comment this link https://github.com/buildbot/buildbot/blame/master/master/contrib/Dockerfile#L61

New web plugin idea: a health indicator

This ticket is a migrated Trac ticket 2966

People contributed to the original ticket: benoit.allard@..., @tardyp, @rutsky, @unknown_contributor, @sa2ajj, @rutsky
Ticket created on: Oct 24 2014
Ticket last modified on: Mar 19 2016

I like to extract as much useful indicator from my builds as possible (time, but also amount of warnings, and such ...)

It would be cool to have a web plugin that could print the evolution of my indicators over time ! (Of course, I would have to configure which indicator I want to see plotted, maybe the kind of plot, and so on ...)

Comment from: @sa2ajj
Date: Oct 24 2014

Could you please elaborate or provide a more specific example?

I think it's related to the [[metrics support|http://docs.buildbot.net/latest/developer/metrics.html]], but without an example I can easily be wrong :)

Comment from: @ben
Date: Oct 24 2014

I was more aiming at [[Statistics|http://docs.buildbot.net/latest/developer/cls-buildsteps.html?highlight=statistics#buildbot.process.buildstep.[[BuildStep]].hasStatistic]], but I just realized that

Note that statistics are not preserved after a build is complete.

So metrics is probably where we want to interface with the master.

I used to abuse Properties for that purpose ...

Comment from: @tardyp
Date: Oct 24 2014

Buildbot plugin system is really made for enabling such dashboards.

A web ui plugin is not technically restricted to creating a bunch of js file, it could also create a twisted service.

For me having the JS only use existing data api to query the data will be very inefficient. I think we could easily create a service, like a status service that registers to a bunch of mq events, and create statistics based on that.

I also had in mind that plugins could have some table in the DB they could use to store the data, or maybe use another db configuration with all schema + migration stuff separated.

Comment from: @tardyp
Date: Oct 26 2014

on IRC, sa2ajj talked about http://graphite.readthedocs.org/

He told us, he actually uses it at work, and has integration working with eigth.

Looking at the documentation, the first reaction is how to integrate this with multimaster, as graphite has its own db called whisper.
I haven't look too much deeply, but I think this is still feasible as a external tool. Probably this would be much cheaper than making our own metrics system inside buildbot.

An external graphite server could be setup, and watch for the (TBD) mq server. As there are messages for nearly every kind of activity that happens in buildbot, this is a good mean of making solid analysis of what is going on. Of course, this solution would not be fully integrated, as probably the UI would be external, but anyway, I think this is a possible cheap path.

@sa2ajj do you think it is possible? How would you estimate the cost of integration?

Comment from: @ben
Date: Oct 26 2014

There are a lot of [[interfaces|http://graphite.readthedocs.org/en/latest/tools.html#visualization]], as well as data collectors / forwarders (same page, a bit higher) available for graphite. It looks like some of them are js-only ! Still way to go for a www-plugin !

Comment from: @sa2ajj
Date: Oct 27 2014

What I said was that we indeed use graphite, but I did not say we use it with Buildbot.

I have an oldish branch where I tried to publish metrics to graphite. However I stopped working on that when I realised that it's not very straightforward to implement it to support multi-master case.

If there's an interest, I can revive the branch or, at least, publish what I have (after rebasing the latest master) so other could comment in what direction my thinking worked.

Comment from: @unknown_contributor
Date: Mar 19 2016

+1 for this feature. I suggest to merge the statistic API with the [[stats module|https://github.com/buildbot/buildbot/tree/master/master/buildbot/statistics]] developed last year by my GSOC student. And add a default backend that store a subset of these stat in the main database that would enable the development of heath/stats visualization modules directly installed by default (using Highchart js lib for example)

[SOLVED]>[question/problem] looks like buildslave not using http_proxy on freebsd.

Have problem with make buildslave use a proxy connection is any tips recommendation available for use buildslave through proxy ? Using freebsd10...

Error what give log is " failed: [Failure instance: Traceback (failure with no frames): <class 'twisted.internet.error.ConnectError'>: An error occurred while connecting: 65: No route to host."

Other programs like pkg/lynx working w/o problems...

buildslave --version
Buildslave version: 0.8.8
Twisted version: 13.2.0

Documentation fails to build with Sphinx >=1.2.2

Documentation fails to build with Sphinx >=1.2.2 due to some fixes in handling of paths.
(See Sphinx issue 1352, Sphinx pull request 211 and these commits: c2917a7a5fe917031c8aebee35c2cbfecc4e3e74, 542b090c2fed1f98d78d2a9764707b7e44ce7384.)

conf.py of Buildbot sets html_favicon = 'buildbot.ico', but buildbot.ico is actually in _static subdirectory.

$ make html
sphinx-build -b html -d _build/doctrees  -q -W . _build/html

Warning, treated as error:
WARNING: favicon file 'buildbot.ico' does not exist

make: *** [html] Error 1

Fix:

--- docs/conf.py
+++ docs/conf.py
@@ -134,7 +134,7 @@
 # The name of an image file (within the static path) to use as favicon of the
 # docs.  This file should be a Windows icon file (.ico) being 16x16 or 32x32
 # pixels large.
-html_favicon = 'buildbot.ico'
+html_favicon = os.path.join('_static', 'buildbot.ico')

 # Add any paths that contain custom static files (such as style sheets) here,
 # relative to this directory. They are copied after the builtin static files,

Gitlab hook is receiving commits in the wrong order

This ticket is a migrated Trac ticket 2886

People contributed to the original ticket: @djmitche, @unknown_contributor, @sa2ajj
Ticket created on: Sep 05 2014
Ticket last modified on: Feb 03 2015

When I receive a push on the Gitlab hook with multiple commits, they are processed in the wrong order and buildbot ends up building a revision that is not the latest.

In the JSON Gitlab is sending, the last commit is the first in the array; it looks like Buildbot expects the reverse.

Thanks!

re-implement progress

This ticket is a migrated Trac ticket 2950

People contributed to the original ticket: @djmitche
Ticket created on: Oct 19 2014
Ticket last modified on: Oct 19 2014

In #2818, I'm removing support for progress as implemented via the status API. It should be re-added!

ETA needed a lot of work, anyway -- see

[[TicketQuery(order=priority,status!=closed,keywords~=eta,format=table,col=summary|owner)]]

BuildSlave keepalive_interval arg is ignored

This ticket is a migrated Trac ticket 2812

People contributed to the original ticket: @djmitche, @sa2ajj, @tomprince
Ticket created on: May 26 2014
Ticket last modified on: Feb 08 2015

In nine, we seem to be ignoring this argument -- at least, a quick grep suggests we are. Perhaps this was part of the landing of the slave-proto stuff?

Comment from: @sa2ajj
Date: Nov 25 2014

Are we talking about [[BuildSlave]] class or we are talking about build slave?

Comment from: @dustin
Date: Feb 08 2015

master/buildbot/buildslave/protocols/pb.py:    # TODO: configure keepalive_interval in c['protocols']['pb']['keepalive_interval']

Handling test results

This ticket is a migrated Trac ticket 2706

People contributed to the original ticket: @djmitche, jonas.pommerening+foss@...
Ticket created on: Feb 26 2014
Ticket last modified on: Jun 18 2014

Me again

I'm currently working on integrating our test results with the [[BuildBot]] web-frontend and mail notifications. I discovered the "halfway there" implementation consisting of [[BuildStatus]].add[[TestResult]] some unfinished web views etc. and thought now would be a good opportunity to tackle that problem.

I already whipped up a little proof-of-concept based on 0.8.8 and am now looking to clean up some rough edges and learn how make the right extensions to nine's the data api (?).
First: What's there already?

[[I[[TestResult]]|https://github.com/buildbot/buildbot/blob/master/master/buildbot/interfaces.py#L595]]: Everything that is known about a single test case: Its name (a tuple of strings), result, a string/text further describing the result, a dictionary (string->string) of logs.

[[I[[BuildStatus]].get[[TestResults]]|https://github.com/buildbot/buildbot/blob/master/master/buildbot/interfaces.py#L577]]: Returns all test results for the given build, keeping I[[TestResult]] per name (where name is I[[TestResult]].getName()).

[[Status[[ResourceBuildTest]]|https://github.com/buildbot/buildbot/blob/master/master/buildbot/status/web/tests.py]]: a web view that can display the result and logs of a single test case.
What do I want to contribute?

I[[LogObserver]] implementations for incrementally parsing test result logs
- [[SubunitLogObserver]] already there
- J[[UnitLogObserver]] parse incrementally with defusedxml (I'll port my lxml-based implementation)
- TAP[[LogObserver]] can someone point me to a python library for that?
- are there any other wide spread formats?
I want to replace the dictionary with a list (possible upgrade hazard?)
- currently the test results are in whatever order the dictionary deems right, the list would assure chronological order
- Allows duplicate names, I'll have to see how to handle that in the ui. I actually think this is needed for data-driven tests (a single method, multiple datasets that are tested within the same test, multiple results).
A "click-through/drill-down" web interface to navigate test results on 0.8.x (#1794)
- linked on the build-status page in the "Result" section with number of passed and failed tests (#536)
- tests results grouped by common prefix of the name tuple (one page per "level")
- maybe a "proportionally split bar" style visualization of failed vs. passed tests
hook it up in nine
- I have no idea what that means, can someone point me in the right direction?

(each of these steps may come as a separate pull request on github)
Why is this important?

This may not yet be the final solution but I think it is a step in the right direction. Buildbot is all about CI and proper integration of test tools should be a big part of that. Right now the whole test result thing is kind of in stealth mode I didn't know about it until last week, despite having worked with Buildbot for about 2 years. The changes I want to make would show that there is a feature like that and maybe inspire discussion and ideas for improving it.
What feedback do I need?

Can I go ahead and change get[[TestResults]] to return a list?
What changes are needed for nine?
Do you have any suggestions, things I should consider, etc?

Comment from: @dustin
Date: Feb 26 2014

This is a great idea!

The existing code for handling test results is, as far as I know, completely inoperative. So you should feel free to throw that out.

The big danger here is an explosion of data. If you have thousands of builds (as most users do) and each build has thousands of tests (as most test runs do), that's millions of test-result rows. It's a lot more for a larger-than-average installation. If we're sending a message for each new test result, that's also a lot of messages. We've done some work with log handling in nine to try to reduce this (by collapsing multiple log lines into a single database record, and compressing it).

I don't think there's a lot of sense in implementing this in 0.8.x, unless you're planning to simultaneously re-implement it in nine. At this point the branches have diverged significantly, and I can't promise to re-implement any new 0.8.x functionality in nine. This would be a substantial rewrite in nine: it will need DB connector plus tables (and new DB migrations in 0.8.x are forbidden..), plus a data API connector, methods for tests to add new test results (which will need to be asynchronous to support DB inserts), and an AngularJS implementation of the frontend.

A good place to start looking at developing for nine is in the developer section of the nine docs - see http://docs.buildbot.net/nine/

Comment from: @unknown_contributor
Date: Feb 26 2014

Some good points!

If you have thousands of builds (as most users do)
check.
and each build has thousands of tests (as most test runs do),
check
that's millions of test-result rows.
ouch so what to do about it? If anything is good at handling that amount of equally structured data it should be a database, right?

It's a lot more for a larger-than-average installation. If we're sending a message for each new test result, that's also a lot of messages.
We've done some work with log handling in nine to try to reduce this (by collapsing multiple log lines into a single database record, and compressing it).

This, indeed, sounds problematic. But wait, you're storing the logs in the database? And you were doing it line-by-line before? Also messages is that related to nine's MQ stuff?

My impression was that it worked something like this:

Slave watches logfile
Slave sends chunks to the master
[[LogObservers]] get updated with received chunks
Master pipes chunks to a file
Master compresses logfile, when step is done
file is read from disk and decompressed when it's needed

But before I'll bother you with any more misinformed assumptions I'll have a look at the nine docs ;)

Comment from: @dustin
Date: Feb 26 2014

In 0.8.x, logs are stored as netstrings in the filesystem, with one file per log. The naive way to port that to the DB, and still allow per-line access, would be to put each line in a DB row. And yes, messages are related to nine's MQ stuff.

If this test support is optional, meaning that users won't find themselves surprised by a 20GB sqlite DB and a slow master, then it might be OK to go ahead and implement this in a fairly naive fashion, maybe a table with columns (buildid, testname, result, logid, logline). If that proves inefficient, we can revisit and optimize.

master does not reconnect to PostgreSQL when database server is restarted

This ticket is a migrated Trac ticket 2753

People contributed to the original ticket: michael+buildbot@..., @djmitche
Ticket created on: Apr 09 2014
Ticket last modified on: Jun 18 2014

On every PostgreSQL upgrade, I need to restart buildbot. I dont think this should be necessary instead, buildbot should simply reconnect :).

Even worse is that the failure mode is pretty horrible. The git_buildbot.py script doesnt output anything and you simply dont get any new builds.

Here is a stack trace of the problem:

2014-04-09 22:19:48+0200 [Broker,7,79.140.39.201] perspective_addChange called
2014-04-09 22:19:48+0200 [-] Peer will receive following PB traceback:
2014-04-09 22:19:48+0200 [-] Unhandled Error
        Traceback (most recent call last):
          File "/usr/lib/python2.7/threading.py", line 783, in __bootstrap
            self.__bootstrap_inner()
          File "/usr/lib/python2.7/threading.py", line 810, in __bootstrap_inner
            self.run()
          File "/usr/lib/python2.7/threading.py", line 763, in run
            self.__target(*self.__args, **self.__kwargs)
        --- <exception caught here> ---
          File "/usr/lib/python2.7/dist-packages/twisted/python/threadpool.py", line 191, in _worker
            result = context.call(ctx, function, *args, **kwargs)
          File "/usr/lib/python2.7/dist-packages/twisted/python/context.py", line 118, in call[[WithContext]]
            return self.currentContext().call[[WithContext]](ctx, func, *args, **kw)
          File "/usr/lib/python2.7/dist-packages/twisted/python/context.py", line 81, in call[[WithContext]]
            return func(*args,**kw)
          File "/usr/lib/python2.7/dist-packages/buildbot/db/pool.py", line 184, in __thd
            rv = callable(arg, *args, **kwargs)
          File "/usr/lib/python2.7/dist-packages/buildbot/db/users.py", line 40, in thd
            rows = conn.execute(q).fetchall()
          File "/usr/lib/python2.7/dist-packages/sqlalchemy/engine/base.py", line 720, in execute
            return meth(self, multiparams, params)
          File "/usr/lib/python2.7/dist-packages/sqlalchemy/sql/elements.py", line 317, in _execute_on_connection
            return connection._execute_clauseelement(self, multiparams, params)
          File "/usr/lib/python2.7/dist-packages/sqlalchemy/engine/base.py", line 817, in _execute_clauseelement
            compiled_sql, distilled_params
          File "/usr/lib/python2.7/dist-packages/sqlalchemy/engine/base.py", line 947, in _execute_context
            context)
          File "/usr/lib/python2.7/dist-packages/sqlalchemy/engine/base.py", line 1108, in _handle_dbapi_exception
            exc_info
          File "/usr/lib/python2.7/dist-packages/sqlalchemy/util/compat.py", line 185, in raise_from_cause
            reraise(type(exception), exception, tb=exc_tb)
          File "/usr/lib/python2.7/dist-packages/sqlalchemy/engine/base.py", line 940, in _execute_context
            context)
          File "/usr/lib/python2.7/dist-packages/sqlalchemy/engine/default.py", line 435, in do_execute
            cursor.execute(statement, parameters)
        sqlalchemy.exc.[[OperationalError]]: ([[OperationalError]]) SSL connection has been closed unexpectedly
         'SELECT users_info.uid \nFROM users_info \nWHERE users_info.attr_type # %(attr_type_1)s AND users_info.attr_data %(attr_data_1)s' {'attr_data_1': u'Michael Stapelberg <[email protected]>', 'attr_type_1': u'git'}

Let me know if you need any other information. Thanks.

Comment from: @dustin
Date: Apr 17 2014

It looks like the postgres dialect needs automatic-reconnection functionality similar to that already in place for MySQL.

http://docs.sqlalchemy.org/en/rel_0_8/core/pooling.html#disconnect-handling-pessimistic

buildmaster doesn't resume when slave unpauses

This ticket is a migrated Trac ticket 2931

People contributed to the original ticket: vlovich@..., @djmitche
Ticket created on: Oct 08 2014
Ticket last modified on: Oct 30 2014

If a slave unpauses, the buildmaster doesn't resume:

Pause the slave. Wait for no jobs being queued on the builder/buildslave.
Submit a job.
Unpause the slave

Expected: job continues running.
Actual: job remains paused until another job is submitted or the buildmaster is reconfig'ed.

Comment from: @dustin
Date: Oct 30 2014

It sounds like the fix here is for the unpausing to call maybe[[StartBuildsForSlave]]

Git: fetch all branches before switching to the given revision, if branch=None

This ticket is a migrated Trac ticket 2750

People contributed to the original ticket: @djmitche, @sa2ajj
Ticket created on: Apr 07 2014
Ticket last modified on: Sep 09 2014

See #1074

branch=None is a bit of a degenerate case, referring vacuously to the "default branch", which in Git really means master. However, it's also the default, and there's no reason to make that default fail in the interest of an optimization (not fetching every commit).

Comment from: @sa2ajj
Date: Sep 08 2014

Maybe the meaning of branch=None should be reconsidered:

if revision is not specified -> default branch
if a revision is specified -> fetch that particular revision and set it as the one to work on

Comment from: @dustin
Date: Sep 09 2014

Maybe I misunderstand, but you can't instruct git to fetch a specific revision from a remote repository -- only to fetch some named revision (and its parents).

Comment from: @sa2ajj
Date: Sep 09 2014

Maybe, I misremembered something.

I'll investigate and update the ticket.

Comment from: @sa2ajj
Date: Sep 09 2014

It looks like

git fetch <remote> <SHA-1>:FETCH_HEAD

works just fine.

$ git --version
git version 2.1.0

Comment from: @dustin
Date: Sep 09 2014

Hah, well, I stand corrected!

buildbot.test.integration.test_try_client.Schedulers.test_userpass_wait is flaky

This ticket is a migrated Trac ticket 2762

People contributed to the original ticket: @djmitche, @sa2ajj, @unknown_contributor
Ticket created on: Apr 20 2014
Ticket last modified on: Aug 27 2015

===============================================================================
[ERROR]
Traceback (most recent call last):
Failure: twisted.trial.util.[[DirtyReactorAggregateError]]: Reactor was unclean.
[[DelayedCalls]]: (set twisted.internet.base.[[DelayedCall]].debug = True to debug)
<[[DelayedCall]] 0x80c0d3b90 [3599.75786614s] called=0 cancelled=0 [[BuildSlave]].doKeepalive()>

buildbot.test.integration.test_try_client.Schedulers.test_userpass_wait
===============================================================================
[ERROR]
Traceback (most recent call last):
Failure: twisted.trial.util.[[DirtyReactorAggregateError]]: Reactor was unclean.
Selectables:
<Broker #0 on 10145>

buildbot.test.integration.test_try_client.Schedulers.test_userpass_wait

Disabled in #2748.

Comment from: @dustin
Date: Apr 11 2015

Fixed in #3246.

Comment from: @unknown_contributor
Date: Aug 15 2015

http://blackjml.livejournal.com/23029.html may describe a better fix.

Comment from: @sa2ajj
Date: Aug 26 2015

Should we port this to nine?