Changes between Version 11 and Version 19 of Ticket #51995


Ignore:
Timestamp:
Apr 14, 2017, 12:38:44 PM (8 years ago)
Author:
mojca (Mojca Miklavec)
Comment:

cross some items that have been done, open a few new tickets

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #51995

    • Property Status changed from new to assigned
    • Property Cc ryandesign raimue neverpanic larryv MarcusCalhoun-Lopez added; ryandesign@… raimue@… cal@… larryv@… removed
    • Property Owner changed from admin@… to admin@…
    • Property Version changed from 2.3.4 to
    • Property Keywords buildbot added
  • Ticket #51995 – Description

    v11 v19  
    11I'm opening a single ticket to list and brainstorm ideas for further improvements of the new buildbot setup. Please note that each individual item listed here would deserve its own ticket, but it makes no sense to open so many tickets at this early stage.
    22
    3 * avoid activating the port and its dependencies if the port was already installed (and is only deactivated), just continue from the upload step
    4 * create "fail cache" and don't build the port if it hasn't been changed and is known to fail; don't even install any dependency if one of the dependencies is know to fail and hasn't been modified since (the email about build failure should still be sent though)
     3* (DONE, [changeset:62834a4212c5b1c59df22221d896cb2c4aab3b44/mpbb 62834a42]) ~~avoid activating the port and its dependencies if the port was already installed (and is only deactivated),~~
     4  * but continue from the upload step (partially addressed by [changeset:4162046ce0f29867d71cca6778d65dbb39b37d05/mpbb 4162046c], still a problem for `wine`)
     5* ~~create "fail cache" and don't build the port if it hasn't been changed and is known to fail; don't even install any dependency if one of the dependencies is know to fail and hasn't been modified since (the email about build failure should still be sent though)~~
    56* provide a force flag to override the cache and attempt the build despite of cache predicting the failure
    6 * provide means to build just a single port without any subports (for example, allow building just `wxWidgets-3.0` without `wxgtk-3.0` and others)
    7 * on top of the `portlist` (`portname`) variable, introduce a new variable that holds variants (say `+quartz -x11`) and allow building the port with those variants (the idea is not to come up with a complete list of useful builds, but to set the necessary building blocks to allow building quartz variants of ports or maybe other tricks that might occasionally be needed); see also #35897
    8 * add a new variable to allow running the tests (we could later try to enable that by default for perl modules for example)
    9 * triggering the individual builds on the builder is currently in random order (not sure if this is possible to fix)
    10 * provide a variable to allow rebuilding all the ports committed between two commit numbers (useful to overcome problems due to temporary broken dependencies or broken build slaves); currently this is only possible for a single commit
    11 * Would it make any sense to merge portwatcher requests or rather not? The slight optimization might reduce the need to rebuild ports with several commits one after the other.
     7* (see see #53976) provide means to build just a single port without any subports (for example, allow building just `wxWidgets-3.0` without `wxgtk-3.0` and others)
     8* (see #35897, #52742) on top of the `portlist` (`portname`) variable, introduce a new variable that holds variants (say `+quartz -x11`) and allow building the port with those variants (the idea is not to come up with a complete list of useful builds, but to set the necessary building blocks to allow building quartz variants of ports or maybe other tricks that might occasionally be needed)
     9* (see #53977) add a new variable to allow running the tests (we could later try to enable that by default for perl modules for example)
     10* (see #52766) triggering the individual builds on the builder is currently in random order (not sure if this is possible to fix)
     11* (see #53978) provide a variable to allow rebuilding all the ports committed between two commit numbers (useful to overcome problems due to temporary broken dependencies or broken build slaves); currently this is only possible for a single commit
     12* (see #53979) Would it make any sense to merge portwatcher requests or rather not? The slight optimization might reduce the need to rebuild ports with several commits one after the other.
    1213
    1314==== Display / UI ====
     
    9899
    99100==== Organisation ====
    100 * move the old `buildbot` and `mpab` to a place of historic interest and rename `buildbot-test` to `buildbot`, update `contrib/README` to reflect the changes
     101* ~~move the old `buildbot` and `mpab` to a place of historic interest and rename `buildbot-test` to `buildbot`, update `contrib/README` to reflect the changes~~
    101102* potentially move more logic of `master.cfg` to a python module
    102 * move the list of builders from hardcoded python code to a configuration file
     103* ~~move the list of builders from hardcoded python code to a configuration file~~