keg: create symlinks in opt for formula aliases #1192

ilovezfs · 2016-09-30T10:23:30Z

No description provided.

ilovezfs · 2016-09-30T11:01:25Z

based on that test run, it sure looks like it will need to be bootstrapped with the existing ones rebuilt against the versioned name

MikeMcQuaid · 2016-09-30T12:04:53Z

CC @Homebrew/maintainers for thoughts on this change and @penman too.

zmwangx · 2016-09-30T12:09:26Z

Could someone write up the rationale behind this change? I see discussions in team chat but not everyone read transcripts after the fact, and @alyssais (looks like the handle changed) doesn't have access to that.

ilovezfs · 2016-09-30T12:16:17Z

The existing bottles are linked against opt/boost not opt/[email protected]. So renaming boost to [email protected] breaks them. The alternatives are not to rename boost to [email protected] and just to go from boost to [email protected] (that pattern is already underway with openssl), or have duplicate boost and [email protected] formulae alongside [email protected], until dependents are transferred from boost to [email protected] or [email protected], and then delete the duplicate boost formula, and create a boost alias instead.

zmwangx · 2016-09-30T12:20:35Z

Do we want to limit opt linking for aliases to @ formulae only? Having both cowthink and cowsay in opt is pointless (but harmless either).

alyssais · 2016-09-30T12:21:47Z

I like the idea that there's nothing special about @ formulae (that they're just normal aliases), but that could definitely end up not being practical.

ilovezfs · 2016-09-30T12:25:38Z

@alyssais I agree with you. I think we'll want a full-on DSL for this stuff at some point relatively soon hehe

alyssais · 2016-09-30T12:59:50Z

I think this looks good. Is it cool if I make a couple of comments on the code?

ilovezfs · 2016-09-30T13:01:13Z

of course!

based on that brew tests run, I'm 👎 on my own PR though :)

alyssais · 2016-09-30T13:02:36Z

based on that brew tests run, I'm 👎 on my own PR though :)

Ah, sure. I'll wait a bit until that's resolved then. :)

MikeMcQuaid · 2016-09-30T13:07:44Z

I like the idea that there's nothing special about @ formulae (that they're just normal aliases), but that could definitely end up not being practical.

I agree with this.

DomT4 · 2016-09-30T13:09:34Z

The alternatives are not to rename boost to [email protected] and just to go from boost to [email protected] (that pattern is already underway with openssl)

This would be my recommendation, FWIW.

DomT4 · 2016-09-30T13:12:11Z

Although we need to resolve the issues around keg_only and so on either way with Boost, especially if my comments on boost and boost-python last night prove accurately remembered. It didn't matter for openssl and [email protected] because both are always going to be keg_only when Homebrew is located in /usr/local.

MikeMcQuaid · 2016-09-30T13:23:00Z

The alternatives are not to rename boost to [email protected] and just to go from boost to [email protected] (that pattern is already underway with openssl)

This would be my recommendation, FWIW.

Even if that's the case I think this alias change would be a good one.

DomT4 · 2016-09-30T13:25:38Z

Yes, I have no issues with aliases also being in opt pointing to the "true" source of the alias, even for non-@ formulae. Might be a good idea to add some tests, if possible, for this change though to ensure we don't bork it in future somehow.

MikeMcQuaid · 2016-10-01T14:38:12Z

Library/Homebrew/keg.rb

Can do the each even if empty? as it'll be a no-op.

MikeMcQuaid · 2016-10-01T14:39:47Z

Library/Homebrew/keg.rb

Can get this from Formula#aliases (which also handles the no-tap case)

MikeMcQuaid · 2016-10-01T14:40:27Z

Library/Homebrew/keg.rb

Again, aliases.each is a no-op if this is empty so you can just use .each without the check.

MikeMcQuaid · 2016-10-01T14:41:35Z

Library/Homebrew/keg.rb

Will a symlink? not always exist??

Broken symlinks return false for exist?

ilovezfs · 2016-10-03T09:01:30Z

@MikeMcQuaid I'm not sure it matters but do you prefer opt/boost -> [email protected] or opt/boost -> ../Cellar/[email protected]?

MikeMcQuaid · 2016-10-03T09:04:06Z

@ilovezfs Then latter just in case people are relying on being able to follow one layer of symlinks to find the Cellar 👍

MikeMcQuaid · 2016-10-03T18:10:11Z

There's some test failures but code and local testing look good to me.

MikeMcQuaid · 2016-10-15T10:46:25Z

@ilovezfs Thoughts on this? Shout if you want help with test failures.

in case people expect to be able to find the prefix by only resolving the symlink once (e.g., if they're using readlink not realpath)

ilovezfs · 2017-02-21T12:44:08Z

@MikeMcQuaid this has introduced a bug for anything that has an alias but was installed before this PR shipped:

iMac-TMP:~ joe$ brew uninstall ag
Uninstalling /usr/local/Cellar/the_silver_searcher/1.0.2... (10 files, 113.1K)
Error: No such file or directory - /usr/local/opt/ag

MikeMcQuaid · 2017-02-21T18:11:56Z

@ilovezfs Working on it.

MikeMcQuaid · 2017-02-21T18:34:11Z

@ilovezfs Opened #2085 to address this.

BrewTestBot added the in progress Maintainers are working on this label Sep 30, 2016

MikeMcQuaid mentioned this pull request Oct 1, 2016

[email protected]: rename from boost Homebrew/homebrew-core#5406

Closed

MikeMcQuaid requested changes Oct 1, 2016

View reviewed changes

ilovezfs closed this Oct 1, 2016

BrewTestBot removed the in progress Maintainers are working on this label Oct 1, 2016

MikeMcQuaid reopened this Oct 1, 2016

MikeMcQuaid added the in progress Maintainers are working on this label Oct 1, 2016

MikeMcQuaid mentioned this pull request Oct 3, 2016

Migrate a Formula to Multiple Versions: add docs. #1204

Merged

5 tasks

antonakv mentioned this pull request Nov 30, 2016

error: the requested upstream branch 'antonakv/cask-repair_update-deadbeef' does not exist vitorgalvao/tiny-scripts#52

Closed

ilovezfs added 2 commits February 20, 2017 14:14

keg: create symlinks in opt for formula aliases

14ef15f

remove unless empty since each is no-op for empty

4c922a0

ilovezfs and others added 2 commits February 20, 2017 14:14

have opt alias symlink point into the Cellar

63295fd

in case people expect to be able to find the prefix by only resolving the symlink once (e.g., if they're using readlink not realpath)

keg: simplify code, handle exceptions.

dfa2c24

MikeMcQuaid force-pushed the optlink_aliases branch from 38e525e to dfa2c24 Compare February 20, 2017 14:23

MikeMcQuaid approved these changes Feb 20, 2017

View reviewed changes

MikeMcQuaid merged commit 60ba0e4 into Homebrew:master Feb 20, 2017

MikeMcQuaid mentioned this pull request Feb 21, 2017

keg: handle missing alias opt link on uninstall. #2085

Merged

Homebrew locked and limited conversation to collaborators May 3, 2018

Uh oh!

keg: create symlinks in opt for formula aliases #1192

keg: create symlinks in opt for formula aliases #1192

Uh oh!

Conversation

ilovezfs commented Sep 30, 2016

Uh oh!

ilovezfs commented Sep 30, 2016

Uh oh!

MikeMcQuaid commented Sep 30, 2016

Uh oh!

zmwangx commented Sep 30, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ilovezfs commented Sep 30, 2016

Uh oh!

zmwangx commented Sep 30, 2016

Uh oh!

alyssais commented Sep 30, 2016

Uh oh!

ilovezfs commented Sep 30, 2016

Uh oh!

alyssais commented Sep 30, 2016

Uh oh!

ilovezfs commented Sep 30, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alyssais commented Sep 30, 2016

Uh oh!

MikeMcQuaid commented Sep 30, 2016

Uh oh!

DomT4 commented Sep 30, 2016

Uh oh!

DomT4 commented Sep 30, 2016

Uh oh!

MikeMcQuaid commented Sep 30, 2016

Uh oh!

DomT4 commented Sep 30, 2016

Uh oh!

MikeMcQuaid Oct 1, 2016

Choose a reason for hiding this comment

Uh oh!

MikeMcQuaid Oct 1, 2016

Choose a reason for hiding this comment

Uh oh!

MikeMcQuaid Oct 1, 2016

Choose a reason for hiding this comment

Uh oh!

MikeMcQuaid Oct 1, 2016

Choose a reason for hiding this comment

Uh oh!

ilovezfs Oct 3, 2016

Choose a reason for hiding this comment

Uh oh!

MikeMcQuaid Oct 3, 2016

Choose a reason for hiding this comment

Uh oh!

ilovezfs commented Oct 3, 2016

Uh oh!

MikeMcQuaid commented Oct 3, 2016

Uh oh!

MikeMcQuaid commented Oct 3, 2016

Uh oh!

MikeMcQuaid commented Oct 15, 2016

Uh oh!

ilovezfs commented Feb 21, 2017

Uh oh!

MikeMcQuaid commented Feb 21, 2017

Uh oh!

MikeMcQuaid commented Feb 21, 2017

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

zmwangx commented Sep 30, 2016 •

edited

Loading

ilovezfs commented Sep 30, 2016 •

edited

Loading