Only check for drive presence
- Queries
- All Stories
- Search
- Advanced Search
- Transactions
- Transaction Logs
Advanced Search
Jul 2 2015
On the server:
Looks like git fetch origin +refs/tags/phabricator/diff/*:refs/tags/phabricator/diff/* is now working... o_o
git fetch origin +refs/tags/phabricator/diff/*:refs/tags/phabricator-diff/*
@epriestley Can we get a DB migration that cleans up all of the bad revisions created by this bug? Apparently it affected a bunch of people and now there are revisions they can't remove from their queue (I can do bin/destroy manually, but that relies on people telling me all of the revision numbers).
Jul 1 2015
I'd suggest just creating Maniphest tasks for high priority items and referencing them in the differential revision with "Ref T1234". Then you can just look at the high priority Maniphest tasks and address them.
@epriestley, does repository mirroring delete branches in the remote target? If not, this could be worked around by just setting the staging repository as a mirror of the original repository (so Phabricator would automatically push commits to the staging repository when master is updated).
Superseded by newer revisions.
Superseded by newer revisions.
Jun 29 2015
Jun 27 2015
I recommend just installing PHP Windows binaries and putting them in your PATH. This + Git for Windows tends to work quite well.
Jun 26 2015
Jun 25 2015
From IRC:
Looks like this has happened on this install a few times as well:
I think we also need a field on a lease to indicate that it's not a permanent lease, or it's a caching lease or something of that nature. Basically the host lease that the working copy has for it's cache shouldn't prevent the host resource from being closed if shouldCloseUnleasedResource returns true.
php -r '$content = file_get_contents("/tmp/diff"); $content = iconv("UTF-8", "UTF-8//IGNORE", $content); echo $content;' | python -c 'import json,sys; print json.dumps(sys.stdin.read())'Yeah, the diff contains invalid UTF8:
Well, I think we're in a bad position, because this code:
to line 27 of AphrontJSONResponse (right after it constructs the JSON), shows:
to line 147 of PhabricatorConduitAPIController shows that the diff content does pass through this line at least once.
Adding file_put_contents('/tmp/diff3', $result); on line 92 of PhabricatorConduitAPIController.php shows that the result is still correctly stored at this point.
at line 136 of DiffusionQueryConduitAPIMethod on the storage tier results in the file having the diff content, so the issue must be further on than this.
When I change line 119 of DiffusionQueryConduitAPIMethod.php to:
I'm like 95% sure this has something to do with the content of the actual diff returned, like UTF-8 encoding or something like that...
Something is totally whack with the cluster configuration in the latest HEAD. For some reason, the storage tier is now constructing a Conduit client to code.pageuppeople.com, even though it should just run the command locally. Even worse, the $is_cluster_request variable is set to true, yet it still constructs a new Conduit client.
For some reason, the storage tier is attempting to forward the request on to another machine inside the cluster (presumably itself?) and is then recursing. This makes no sense given that the code in src/applications/diffusion/conduit/DiffusionQueryConduitAPIMethod.php at line 123 is designed to prevent recursion.
I ran the query directly against the Conduit API on the storage machine, from my local machine, and it returns an empty response (so it's got nothing to do with the requesting machine):
I modified diffusion.rawdiffquery to do the following before returning the result:
Jun 24 2015
There is a URI artifact type, so if you're writing your own custom build step in PHP, you can create a URI artifact that points to the CircleCI plan:
Another option is to add the following to .gitattributes:
Remove unrelated changes
Jun 19 2015
Jun 18 2015
Yeah this does already work (I've used it before). The HTTPS URL is just the same URL as the Diffusion page, even when the repository is hosted externally (SSH also works).
I ran chown -R git:wwwgrp-phabricator . in the storage folder which fixed it. Although it is a little concerning that it started happening for no reason (and has been working fine for ages).
Jun 16 2015
Jun 15 2015
Also this doesn't line up:
This looks weird to me because my brain continues the straight line downward and doesn't like the fact that the circular icons rest right against that invisible line. Maybe it's me, but I feel like there should be some left padding on those circular icons...
Jun 14 2015
Jun 12 2015
Jun 9 2015
This is caused by not having the PHP XmlWriter extension installed. On OpenSUSE I had to do:
Jun 8 2015
Jun 4 2015
May 23 2015
I don't believe there's a deadlock here. What happens is:
May 22 2015
"Block and wait" is what currently happens when you have more leases come through than maximum resources available (actually for blueprints that implement min-max-expiry it will optimize to only allocate as many resources as is really needed).
May 19 2015
That sounds similar to the solution I proposed, but I'm not entirely sure that will work due to the health check running as a queued task (and preemptively preventing leases against suspicious resources I guess). I'm particularly concerned about this scenario:
With both saved search and project tagging, this is reasonably solved.
May 16 2015
It's not actually the same repository, they just have common history (3 of the repos are actually archived...)
Oh heh, I searched for "repeat" on the Calendar workboard and couldn't find anything, so I assumed there wasn't a task already.
May 13 2015
Yeah it makes sense in most places, but in Phriction it's the only way to access the parent document.