Difference between revisions of "User:Sardisson/madhatter"

From Camino Wiki
Jump to navigation Jump to search
(→‎BUGS: more info)
(→‎BUGS: symbol upload failures)
 
(2 intermediate revisions by the same user not shown)
Line 6: Line 6:
 
* [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/rev/136a0b3f03ce Show revision info for madhatter self-updates]
 
* [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/rev/136a0b3f03ce Show revision info for madhatter self-updates]
 
* In theory, Hg checkout should support pull-by-date: [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#1257], [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1013]
 
* In theory, Hg checkout should support pull-by-date: [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#1257], [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1013]
* Support different branches/tags in each repo
+
* <s>Support different branches/tags in each repo</s> [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/rev/bf5bfee0d0de]
 
* Improve the changeset generation code and handle failures/errors: [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1350], [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1369]
 
* Improve the changeset generation code and handle failures/errors: [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1350], [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1369]
  
 
== BUGS ==
 
== BUGS ==
 
+
* Need better error visibility when uploading symbols fails (see [https://bugzilla.mozilla.org/show_bug.cgi?id=491101#c8 bug 491101 comment 8])
 +
** We don't want to make the build fail to upload if the build itself is fine (we should probably fail to upload the build if *generating* the symbols fails)
 +
** Since for nightlies we build when (sane) US people are asleep, we don't want to make them have to scroll/page back to find the build/log on tinderbox (although at least TinderPrinting would help there); can we email the build administrator in that case?
 
* (When building from a branch), getting the Gecko revision fails sometimes (Cm2.1b1)
 
* (When building from a branch), getting the Gecko revision fails sometimes (Cm2.1b1)
 
** One option might be to get the info from <code>platform.ini</code>/<code>generated/*_changeset</code> cache files in the Camino build, if they're empty.  However, currently we do the revision-generation code after cloning rather than at TinderPrint time (so we know where we are in the filesystem), and those files don't exist before a build.
 
** One option might be to get the info from <code>platform.ini</code>/<code>generated/*_changeset</code> cache files in the Camino build, if they're empty.  However, currently we do the revision-generation code after cloning rather than at TinderPrint time (so we know where we are in the filesystem), and those files don't exist before a build.
Line 16: Line 18:
 
** Especially useful when cvs.mozilla.org gets misconfigured and is inaccessible; madhatter just appears to stop building while it waits for cvs.m.o to become available again, but there's no log info at all.
 
** Especially useful when cvs.mozilla.org gets misconfigured and is inaccessible; madhatter just appears to stop building while it waits for cvs.m.o to become available again, but there's no log info at all.
 
* Clobber of mozilla sometimes fails: [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1276]
 
* Clobber of mozilla sometimes fails: [http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1276]
* Failures cloning mozilla don't kill the build at that point (related to above?): [<!--http://tinderbox.mozilla.org/showlog.cgi?tree=Camino&errorparser=unix&logfile=1311836580.1311837330.10857.gz&buildtime=1311836580&buildname=MacOSX%20Darwin%209.8.0%20cb-xserve04%20Depend%20Cm2.1-M1.9.2&fulltext=1-->http://www.ardisson.org/smokey/moz/1311836580.1311837330.10857.gz tinderbox log]
+
<ul><li> Failures cloning mozilla don't kill the build at that point (related to above?): [<!--http://tinderbox.mozilla.org/showlog.cgi?tree=Camino&errorparser=unix&logfile=1311836580.1311837330.10857.gz&buildtime=1311836580&buildname=MacOSX%20Darwin%209.8.0%20cb-xserve04%20Depend%20Cm2.1-M1.9.2&fulltext=1-->http://www.ardisson.org/smokey/moz/1311836580.1311837330.10857.gz tinderbox log]
** Related?  {{bug|330779}}
+
<ul><li> Related?  {{bug|330779}}
* The "status" returned from the run_shell_command_with_timeout() for checkout is wrong or missing or something.[http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1300] For 2.1rc3 attempt 1, instead of the expected "Error: Hg checkout timed out.", the log reports:
+
<li> The "status" returned from the run_shell_command_with_timeout() for checkout is wrong or missing or something.[http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1300]<br>For 2.1rc3 attempt 1, instead of the expected "Error: Hg checkout timed out.", the log reports:
<pre>Begin: Sun Nov 13 19:00:16 2011
+
<br><pre>Begin: Sun Nov 13 19:00:16 2011
 
hg clone -r CAMINO_2_1_RELEASE http://hg.mozilla.org/releases/mozilla-1.9.2/ mozilla  
 
hg clone -r CAMINO_2_1_RELEASE http://hg.mozilla.org/releases/mozilla-1.9.2/ mozilla  
 
Process killed. Took 2 seconds to die.
 
Process killed. Took 2 seconds to die.
End:  Sun Nov 13 19:10:17 2011</pre>
+
End:  Sun Nov 13 19:10:17 2011</pre>That error comes from <code>kill_process()</code>[http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1901], called once we believe we've timed out, so it's unclear how build_status is not getting set to busted.</li></ul></ul>
: That error comes from <code>kill_process()</code>[http://hg.mozilla.org/users/alqahira_ardisson.org/madhatter/file/tip/build-seamonkey-util.pl#l1901], called once we believe we've timed out, so it's unclear how build_status is not getting set to busted.
 

Latest revision as of 23:55, 30 April 2012

madhatter is a fork of the Tinderbox client that supports building from either CVS or Hg. It was hacked up by Smokey Ardisson to support Camino 2.1 development.

TODO

Usually marked XXX in the code, although some of them are from old tinderbox hackers; this list is not complete

BUGS

  • Need better error visibility when uploading symbols fails (see bug 491101 comment 8)
    • We don't want to make the build fail to upload if the build itself is fine (we should probably fail to upload the build if *generating* the symbols fails)
    • Since for nightlies we build when (sane) US people are asleep, we don't want to make them have to scroll/page back to find the build/log on tinderbox (although at least TinderPrinting would help there); can we email the build administrator in that case?
  • (When building from a branch), getting the Gecko revision fails sometimes (Cm2.1b1)
    • One option might be to get the info from platform.ini/generated/*_changeset cache files in the Camino build, if they're empty. However, currently we do the revision-generation code after cloning rather than at TinderPrint time (so we know where we are in the filesystem), and those files don't exist before a build.
  • Need to have self-update and tinder-config update generate a log so someone can at least check locally what's going on when they fail. Ideally, use the next build's log, but that's not set up yet, and return the purple (infra exception) status.
    • Especially useful when cvs.mozilla.org gets misconfigured and is inaccessible; madhatter just appears to stop building while it waits for cvs.m.o to become available again, but there's no log info at all.
  • Clobber of mozilla sometimes fails: [6]
  • Failures cloning mozilla don't kill the build at that point (related to above?): tinderbox log
    • Related? Bug 330779
    • The "status" returned from the run_shell_command_with_timeout() for checkout is wrong or missing or something.[7]
      For 2.1rc3 attempt 1, instead of the expected "Error: Hg checkout timed out.", the log reports:
      Begin: Sun Nov 13 19:00:16 2011
      hg clone -r CAMINO_2_1_RELEASE http://hg.mozilla.org/releases/mozilla-1.9.2/ mozilla 
      Process killed. Took 2 seconds to die.
      End:   Sun Nov 13 19:10:17 2011
      That error comes from kill_process()[8], called once we believe we've timed out, so it's unclear how build_status is not getting set to busted.