Permalink
Find file
Fetching contributors…
Cannot retrieve contributors at this time
109 lines (90 sloc) 5.04 KB
thrown into a shark poind: surviving core development
speaker: gabor hojtsy
drupalcon 2012
shiny handout: "I have an idea for Drupal core!"
gabor is maintainer for d6 (former?), works for acquia, D8MI (mobile initiative)
step 0: decide which drupal version you want to add a feature to. d8 gets new shiny, d7 gets mostly backports and small changes, d6 is maint. only
d7 has a looser backporting policy than d6 did, so there's some potential for increased awesome
d6 needs to stay super-stable. shiny doesn't happen there anymore, though perf improvements are welcome
d6 definitely gets security fixes
to fix a d6 issue, patch d8 first, backport to d7, backport to d6. wheee.
this is good for users, but kinda sucks for hackers
there are no forward ports because bug fixed in version X need to stay fixed after version X
step 1: Do you need it done right or now?
need it right now: hack it (yuck), use hooks, use contrib modules
even if your patch won't get into core, submit an issue to core anyway
drush make helps you with this
you can still get some of the benefits of the oss community (review, validation), even if you don't get all of them
override hooks:
hook_menu_alter() lets you mess with routing arbitrarily
hook_module_implements_alter() lets you mess with how module hooks are invoked arbitrarily (?!?!)
using these implies some copy/pasting
hook_form_alter, hook_page_alter - mess with forms and page output arbitrarily
be careful. internal apis might change.
bigger changes imply more divergence over time
post a contrib module if possible
always be careful with these. with great power...
propose d8 changes if possible
need it done the right way: work on d8
this is preferred
you get feedback from the community
help you use best practices, use the right api, use the right approach
easier to get fixes across different drupal version
it's not solely your responsibility
other people get to build off your changes
getting it into d8 core:
dont' just start an issue. there are too many already without some differentiating feature
start a conversation unless it's a small/obvious fix
manage your patch throughout the review process
likely first reaction: nobody cares at all. there are 9500 core issues
avg lifetimes: bug - 33 weeks, feature req - 1y24w, support - 34w, tasks - 43w
ow.
check out issue submission guidelines - d.o node 73179, issue tags on d.o node 1207020
there are drupal core initiatives. there are a handful and they cover big topics.
core initiatives: g.d.o/drupal-initiatives
community initiatives: d.o/community-initiatives/drupal-core (less up-to-date, still useful)
working groups on g.d.o/groups
when working on pushing an issue forward, look at thie git history
get issue numbers
find usernames of hackers who are likely to care about the thing
d.o/irc explains how irc works for the drupal world
d.o/core-office-hours - find when someone's available (?)
d.o/planet - you can blog about how you're working on an initiative or what not
can get some feedback through this, plus blogging is generally a good idea
now, everyone cares about it. OH NOES!
consider all feedback and criticism
don't cater to everyone
keep the issue summary updated
be present and response where needed
this will take some serious time
have an overall plan with followups
don't let the issue get sidetracked. see dcoc for how to deal
see also: d.o/dcoc (drupal code of conduct)
3 approaches to getting stuff into core:
d.o project sandbox
more flexibility, no risk of breaking core before stuff is ready
separate issue queue
CMI uses this (config mgmt initiative)
at merge time, most poeple haven't been patying attention, so there's ramp-up
structure tagging (e.g. "D8MI")
work directly in d.o issue queue
metaissues (issues that list relevant sub-issues)
entity initiative uses this approach
YAY; patch is rtbc!
no guarantees that it'll get into core or not reverted
still need to keep an eye on it and make sure it gets into core
people can still complain about e.g. codingstd or perf problems
once it's in, PARTY!
see also, denver2012.drupal.org/sprints (if you happen to be at drupalcon denver)
q/a:
Q; how do you integrate core hacking into $dayjob's workflow
use acquia dev desktop as dev env (or whatever. nothing weird's needed)
note to self: hack something (or find solution) to quickly spin up a new drupal instance for hacking. manual process sucks atm
xjm: I don't have ideas for drupal. I help other people who have ideas for drupal.
also look for "novice" tagged issues in core
core initiatives hold bi-weekly meetings on irc, so jump on in
if you read a huge issue comment thread, post an issue summary
Q: is there any plan to decrease the issue timeline?
it's hard and most people don't use the dev version, and testing sucks
verifying patches is a highly manual process
latest d6 release fixed two bugs, both of which were introduced in the previous release