User:Rcampbell/1:30 Build Infrastructure notes: Difference between revisions

From MozillaWiki
Jump to navigation Jump to search
No edit summary
 
m (GPHemsley moved page Robcee 1:30 Build Infrastructure notes to User:Rcampbell/1:30 Build Infrastructure notes without leaving a redirect: User page)
 
(4 intermediate revisions by one other user not shown)
Line 1: Line 1:
Test tinderbox starts up, downloads hourly builds and runs tests. VMs don't provide reliable timing data.
[[Build farm infrastructure update and discussion]]


mozilla-tests tree on tinderbox.
[http://tinderbox.mozilla.org/showbuilds.cgi?tree=Firefox]


[[Build farm infrastructure update and discussion]]
paul: mozilla tools tinderbox-configs
 
Update information on the wikis to be more accurate.
 
dbaron: Build targets should be added to client,mk and not restricted to Tinderbox. There is some consensus.
 
make check
 
rhelmer: Tinderbox currently runs "make test" or could run "make perftest"
 
boris/dbaron: performance tests should run on slow boxes, correctness tests should run as quickly as possible. Dbaron notes that correctness tests, if done possible should take quite a while.
 
schrep: can see need for three classes of machines: builds, performance, correctness.
 
schrep: conceivably having a build system analogous to Java+Junit+Ant where a make target could invoke a set of tests and output the results to a text or xml file for Tinderbox to pick up. This allows developers to run tests as well as have them automatically invoked in tinderbox.

Latest revision as of 00:57, 11 May 2015

Build farm infrastructure update and discussion

[1]

paul: mozilla tools tinderbox-configs

Update information on the wikis to be more accurate.

dbaron: Build targets should be added to client,mk and not restricted to Tinderbox. There is some consensus.

make check

rhelmer: Tinderbox currently runs "make test" or could run "make perftest"

boris/dbaron: performance tests should run on slow boxes, correctness tests should run as quickly as possible. Dbaron notes that correctness tests, if done possible should take quite a while.

schrep: can see need for three classes of machines: builds, performance, correctness.

schrep: conceivably having a build system analogous to Java+Junit+Ant where a make target could invoke a set of tests and output the results to a text or xml file for Tinderbox to pick up. This allows developers to run tests as well as have them automatically invoked in tinderbox.