Summary: Fixes T2605. Provide some instructions on configuring RDS properly. The "DB Parameter Group" thing in the web UI seems pretty easy to use, it's just not obvious that it's what you should be using.
Test Plan: Jiggled these warnings to trigger them, viewed the output, saw a table of values and a hint about RDS.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T2605
Differential Revision: https://secure.phabricator.com/D10343
Summary:
Ref T2605. For old MySQL, this option is not supported. Catch that and tailor the error.
I couldn't find the first version of MySQL which introduced this optino in order to produce a more useful error. I spent about ~10 minutes looking.
Test Plan: Faked the error, survived setup.
Reviewers: btrahan, chad
Reviewed By: chad
Subscribers: epriestley
Maniphest Tasks: T2605
Differential Revision: https://secure.phabricator.com/D10342
Summary: Fixes T4881.
Test Plan: made a config change, saw the issue, restarted daemons and it went away
Reviewers: epriestley
Reviewed By: epriestley
Subscribers: epriestley, Korvin
Maniphest Tasks: T4881
Differential Revision: https://secure.phabricator.com/D10339
Summary: Fixes T4130. Adds a setup warning when ft_min_word_len is set to the default value.
Test Plan: Hit setup warning; resovled setup warning. Searched for "DOS".
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T4130
Differential Revision: https://secure.phabricator.com/D10259
Summary:
Fixes T2605.
- Add a setup warning about the stopword file.
- Provide a simpler stopword file.
Test Plan:
- Hit setup warning.
- Resolved it according to instructions.
- Added "various" to a task, then searched for it, found the task.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T2605
Differential Revision: https://secure.phabricator.com/D10258
Summary:
Fixes T5545. We assume `strlen()` returns the number of bytes in a string, which is the normal behavior (and the documented behavior).
There's a config option, `mbstring.func_overload`, which silently calls mb_strlen() instead. This may return some other result, might fail, etc., and there's no way to get the byte length of a string if this option is set.
If this option is set, fatal immediately. Nothing good can ever come of it.
Test Plan: {F173990}
Reviewers: btrahan, chad
Reviewed By: chad
Subscribers: epriestley
Maniphest Tasks: T5545
Differential Revision: https://secure.phabricator.com/D9811
Summary: Ran `arc lint --apply-patches --everything` over rP, mainly to change double quotes to single quotes where appropriate. These changes also validate that the `ArcanistXHPASTLinter::LINT_DOUBLE_QUOTE` rule is working as expected.
Test Plan: Eyeballed it.
Reviewers: #blessed_reviewers, epriestley
Reviewed By: #blessed_reviewers, epriestley
Subscribers: epriestley, Korvin, hach-que
Differential Revision: https://secure.phabricator.com/D9431
Summary: This doesn't look like it is spelled correctly.
Test Plan: Seems more correct now.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Differential Revision: https://secure.phabricator.com/D9043
Summary: There was a typo.
Test Plan: Now there is no typo.
Reviewers: btrahan, chad
Reviewed By: chad
Subscribers: aran, epriestley
Differential Revision: https://secure.phabricator.com/D8995
Summary: Fixes T4916. Although every normal build of PHP has this in the core, at least one distribution which users could reasonably encounter does not.
Test Plan: Changed string to "ctypex", got setup warning. Changed to "ctype", got no warning.
Reviewers: hach-que, btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T4916
Differential Revision: https://secure.phabricator.com/D8896
Summary:
- Point them at the new Diviner.
- Make them a little less cumbersome to write.
Test Plan: Found almost all of these links in the UI and clicked them.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Differential Revision: https://secure.phabricator.com/D8553
Summary: Ref T2222. These no longer have an effect, and are obsoleted by `differential.fields`.
Test Plan: `grep`
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T2222
Differential Revision: https://secure.phabricator.com/D8468
Summary:
A user reported this stack trace:
http://pastebin.com/6auGbZsE
...on this GitHub issue:
https://github.com/facebook/phabricator/issues/389#issuecomment-36612511
The problem is similar to the original report, but not identical. In this case, we're following a sequence of steps like:
- Run setup checks.
- Check for enabled providers, in order to raise "no providers configured yet" warning.
- Try to generate login/redirect URIs.
- Build the request.
- Set the default base URI.
- Run normal code.
Since we try to generate URIs before we provide a default, this fatals. Instead, don't try to build objects.
An alternative fix might be to try to set defaults earlier, but we depend on some config and on building the Request in order to be able to figure out if a request is HTTP or HTTPS right now. We could assume one, or guess, or use protocol-relative URIs (`///host.com`), but I think this fix is a little cleaner overall. If we keep hitting similar stuff, we could look into alternate fixes.
We could also set some kind of "setup mode" flag and make `getURI()` if it's called during setup mode to detect these during testing. I'd like to hit one more of these before doing that, though.
Test Plan: Reproduced the issue, applied the patch, verified this fixes it.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Differential Revision: https://secure.phabricator.com/D8395
Summary: Ref T2222. This will probably have some rough edges for a bit (e.g., weird cases I didn't remember or think of), but there's no change to the underlying data and we can easily revert if things get too messy.
Test Plan: Looked at a variety of revisions and saw sensible output.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T2222
Differential Revision: https://secure.phabricator.com/D8361
Summary:
See D8308. Enabling STRICT_ALL_TABLES prevents this entire class of error, by fataling on truncation instead of truncating. We never want truncation; it is always bad and sometimes extremely bad.
We've recommended this mode for developer installs for a long time, and some users run with it enabled, so it's very unlikely to cause any issues (I've had it enabled locally for at least 6-8 months, I think).
Test Plan:
- Disabled mode.
- Saw warning.
- Enabled mode.
- No warning.
{F117040}
Reviewers: btrahan, chad
Reviewed By: chad
CC: chad, aran, arice
Differential Revision: https://secure.phabricator.com/D8309
Summary:
Ref T4420. Tokenizers currently operate in "preload" or "ondemand" modes. In the former mode, which is default, they'll try to load the entire result list when a page loads.
The theory here was that this would slightly improve the experience for small installs, and once they got big enough they could switch to "ondemand". In practice, several issues have arisen:
- We generally don't have a good mechanism for telling installs that they should tweak perf config -- `metamta.send-immediately` is the canonical example here. Some large installs are probably affected negatively by not knowing to change this setting, and having settings like this is generally annoying.
- We have way way too much config now.
- With the advent of ApplicationSearch, pages like Maniphest make many redundant loads to prefill sources like projects. Most of the time, this data is not used. It's far simpler to switch everything to ondemand than try to deal with this, and dealing with this would mean creating two very complex divergent pathways in the codebase for a mostly theoretical performance benefit which only impacts tiny installs.
- We've been using `tokenizer.ondemand` forever on `secure.phabricator.com` since we have many thousands of user accounts, and it doesn't seem sluggish and works properly.
Removing this config is an easy fix which makes the codebase simpler.
I've retained the ability to use preloaded sources, since they may make sense in some cases (in at least one case -- task priorities -- adding a static source pathway might make sense), and they're part of Javelin itself. However, the code will no longer ever go down that pathway.
Test Plan: Used `secure.phabricator.com` for years with this setting enabled.
Reviewers: btrahan, chad
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T4420
Differential Revision: https://secure.phabricator.com/D8232
Summary:
Ref T4310. Fixes T3720. This change:
- Removes concurrent session limits. Instead, unused sessions are GC'd after a while.
- Collapses all existing "web-1", "web-2", etc., sessions into "web" sessions.
- Dramatically simplifies the code for establishing a session (like omg).
Test Plan: Ran migration, checked Sessions panel and database for sanity. Used existing session. Logged out, logged in. Ran Conduit commands.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T4310, T3720
Differential Revision: https://secure.phabricator.com/D7978
Summary:
Ref T3857.
- Always send mail via daemons. This lets us get rid of this config, and is generally much more performant.
- After D7964, we warn if daemons aren't running.
Test Plan: Sent some mail.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T3857
Differential Revision: https://secure.phabricator.com/D7965
Summary:
Currently, we try to mostly-kind-of-work if daemons aren't running (for example, we send mail in-process). I want to stop doing this. A major motivator is that `metamta.send-immediately` is confusing for a lot of users and frequently the cause of performance problems. Increasingly, functionality of applications depends on the daemons (Harbormaster, Drydock, Nuance all require daemons to do anything at all). They're also fairly stable/robust/well-tested and no reasonable install should be running without them.
This will let us simplify or remove some flags (like `metamta.send-immediately`) and simplify some other processes like search indexing.
Test Plan: Stopped daemons, loaded warnings, saw daemon warning. Started daemons, reloade, no warning.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T3857
Differential Revision: https://secure.phabricator.com/D7964
Summary: Ref T4222. This was used by Facebook while developing Releeph, but should no longer be necessary since Releeph is in the upstream. I can't get an answer out of Facebook about whether they still use it or not (see T4227), so nuke it. We're going to replace it with a more general mechanism (see T4222).
Test Plan: Regenerated celerity map. Browsed some pages, still got resources.
Reviewers: btrahan, hach-que
Reviewed By: hach-que
CC: aran
Maniphest Tasks: T4222
Differential Revision: https://secure.phabricator.com/D7863
Summary:
Fixes T3034. This is obsoleted by modern policies.
This was written by a Facebook intern and is rarely used -- the Hive install might be the only use in the wild. It has never really worked correctly.
Test Plan: `grep`; browsed Differential.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T3034
Differential Revision: https://secure.phabricator.com/D7568
Summary: I'm planning to add more detailed info to Diffusion itself, but catch the big issue here.
Test Plan: Hit config issue locally, then resolved it.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Differential Revision: https://secure.phabricator.com/D7439
Summary: We've had support for this for a long time, but it was conditional on config. Since it more-or-less actually does something now, just enable it unconditionally.
Test Plan: Settings -> SSH Public Keys
Reviewers: btrahan
Reviewed By: btrahan
CC: hach-que, aran
Maniphest Tasks: T2230
Differential Revision: https://secure.phabricator.com/D7426
Summary:
Gets rid of some old Differential-specific nonsense and replaces it with general runtime-pluggable Remarkup rules.
Facebook: This removes two options which may be in use. Have any classes being added via config here just subclass the new abstract bases instead. This should take 5 seconds to fix. You can adjust order by overriding `getPriority()` on the rules, if necessary.
Test Plan: See comments.
Reviewers: btrahan
Reviewed By: btrahan
CC: FacebookPOC, andrewjcg, aran
Differential Revision: https://secure.phabricator.com/D7393
Summary: Ref T3794. Drop auxiliary field, use standard field.
Test Plan: Performed migration, field seemed to survive it intact. Edited and viewed tasks.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T3794
Differential Revision: https://secure.phabricator.com/D7036
Summary: Ref T418. Maniphest has an obsolete class-based field selector. Replace it with CustomField-based selectors, which use the nice config UI and are generally way easier to use.
Test Plan: Added custom fields; edited and viewed custom fields on tasks. Everything worked as expected.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T418
Differential Revision: https://secure.phabricator.com/D6998
Summary: See IRC. This is dumb but I think we should try to work by default on Debian, and it doesn't cost us too much. See inline comment for more.
Test Plan:
- No `disable_functions`, restarted, worked fine.
- Set `disable_functions = pcntl_derp`, restarted, worked fine.
- Set `disable_functions = derp`, restarted, setup fatal.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Differential Revision: https://secure.phabricator.com/D6741
Summary:
Fixes T3709. PHP has two configuration options ('disable_functions', 'disable_classes') which allow functions and classes to be blacklisted at runtime.
Since these break things in an unclear way, raise a setup fatal if they are set.
We take a slightly more tailored approach to these in `phd` already, but I'd rather try just saying "no, this is bad" and see if we can get away with it. I suspect we can, and there's no legitimate reason to blacklist functions given that Phabricator must have access to, e.g., `proc_open()`.
Test Plan: {F54058}
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T3709
Differential Revision: https://secure.phabricator.com/D6739
Summary:
See discussion in T2715. Currently, PHIDs are all hard coded in the PHID application. In the long run, we need to move them out into actual applications.
A specific immediate issue is Releeph, which uses a very very old and very broken mechanism to inject PHIDs in a way that only sort of works.
Moving forward, every PHID type will be provided by a `PhabricatorPHIDType` subclass, which will manage loading it, etc.
This also moves toward cleaning up the "load objects by name" (where "name" means something like `D12`) code, which is an //enormous// mess and spread across at least 4-5 callsites.
Test Plan: Used `phid.lookup` and `phid.query` to load Slowvotes.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Differential Revision: https://secure.phabricator.com/D6502
Summary:
These versions are broken, but package distros seem to be picking them up. :/
Since the error you get is completely useless, fatal immediately with a useful message.
Ref T2594.
Test Plan: Faked verisions and hit the issue.
Reviewers: btrahan
Reviewed By: btrahan
CC: brennantaylor, Arijit, aran
Maniphest Tasks: T2594
Differential Revision: https://secure.phabricator.com/D6415
Summary: Fixes T3501. `apc.stat` should generally be 0 in production and 1 in development. Raise a setup warning if it isn't.
Test Plan:
Hit both setup warnings.
{F49176}
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T3501
Differential Revision: https://secure.phabricator.com/D6376
Summary: Fixes T3400. Users are crafty. Attempt to outwit them.
Test Plan: Added all kinds of nonsense to my PATH to hit all the errors. Verified sensible-looking error messages which I couldn't figure out any way to misread or outwit.
Reviewers: chad, btrahan
Reviewed By: chad
CC: aran
Maniphest Tasks: T3400
Differential Revision: https://secure.phabricator.com/D6318
Summary:
Fixes T3436. Currently, when installs have configuration options which we don't know about, we raise a fairly confusing/ambiguous message about the options being unknown. Instead:
- Keep a list of previously valid (but now deleted) config, with explanatory reasons for what happened to it. Present this information, along with altenate wording ("Obsolete Config" instead of "Unknown Config") where applicable.
- Show a list of all the places the config is defined.
- Provide an active link to delete it from the web UI.
- Provide a command to delete it from the CLI.
- Allow `bin/config delete` to delete configuration options which no longer have a definition.
Test Plan:
- Set an auth key in database, local and file config.
- Walked through the setup issue, cleaning it up.
- Set an invalid key and made sure I still got a reasonable error (this now has better cleanup instructions).
Reviewers: btrahan, chad
Reviewed By: chad
CC: aran
Maniphest Tasks: T3436
Differential Revision: https://secure.phabricator.com/D6317
Summary: Ref T1536. This is the last major migration. Moves us over to the DB and drops all the config stuff.
Test Plan:
- Ran the migration.
- Saw all my old config brought forward and respected, with accurate settings.
- Ran LDAP import.
- Grepped for all removed config options.
Reviewers: btrahan, chad
Reviewed By: btrahan
CC: aran, wez
Maniphest Tasks: T1536
Differential Revision: https://secure.phabricator.com/D6243
Summary:
Ref T1536. Currently, when you install Phabricator you're dumped on the login screen and have to consult the documentation to learn about `bin/accountadmin`.
Instead, detect that an install is running first-time setup:
- It has no configured providers; and
- it has no user accounts.
We can safely deduce that such an install isn't configured yet, and let the user create an admin account from the web UI.
After they login, we raise a setup issue and lead them to configure authentication.
(This could probably use some UI and copy tweaks.)
Test Plan:
{F46738}
{F46739}
Reviewers: chad, btrahan
Reviewed By: chad
CC: aran
Maniphest Tasks: T1536
Differential Revision: https://secure.phabricator.com/D6228
Summary:
Ref T3354. There's no way for us to test most of the config options which actually affect this limit, so the Phabricator config is basically a canary value to indicate "the administrator hasn't configured anything yet".
Raise a setup issue if it isn't set. There's a trail to get here from Files, but we've de-emphasized the old-school upload form so it's hard to unearth.
Emphasize the warning that you need to read the documentation and configure like 30 other things to make this work.
Test Plan: Cleared my config, verified I got the issue, read it, set my config, issue went away.
Reviewers: jamesr, chad
Reviewed By: chad
CC: aran
Maniphest Tasks: T3354
Differential Revision: https://secure.phabricator.com/D6185
Summary:
These are a bit tricky because we don't want to require you to install a VCS you don't use just to use Phabricator. Test that repositories exist before performing the checks.
I'll couple this with additional checks during repository creation.
Test Plan: Changed binary names to nonexistent ones, verified setup issues raised properly.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Differential Revision: https://secure.phabricator.com/D6040
Summary:
Spent an hour or two helping a user figure this out. Make sure I never do that again.
If the webserver is configured with an empty or bogus PATH, binaries like 'which' and 'diff' (and 'git', and 'svn', etc.) may not be available. In most cases, this is fine, because we get an error like "sh: whatever-command not found", which is obvious to diagnose.
In the case of 'diff', we don't get this, because 'diff' is expected to exit with a nonzero code for differing files -- so we interpret the "sh: whatever-command not found" as "files differ" and then try to parse the empty output.
Explicitly check for 'which' (on Windows, 'where') and 'diff' during setup (I plan to refine the behavior around 'git', 'svn' and 'hg' at some point, but this is less pressing since the errors are trivial to support).
Test Plan: Faked failures on all modes, verified setup warnings look reasonable.
Reviewers: btrahan, chad
Reviewed By: btrahan
CC: aran
Differential Revision: https://secure.phabricator.com/D6008
Summary:
See <https://github.com/facebook/phabricator/issues/320>. We have a soft dependency on 'fileinfo', which we try to recover from (with `file`) but won't be able to on Windows and apparently FreeBSD systems. Since users can ignore setup checks anyway now, just raise a warning during install.
I believe almost all installs should have this extension, it has been part of the core for a long time.
Test Plan: Faked setup failure, looked at warning. "Solved" setup failure, saw it go away.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Differential Revision: https://secure.phabricator.com/D5952
