Strange. The failing unit test succeeds on my machine, even with this change checked out.

Have you set a VCS Password (see your user settings) and are you using that password when using arc land?

MariaDB added support for SHOW REPLICA STATUS in MariaDB 10.5.1, which was released February 14, 2020. This means compatibility between MySQL and MariaDB diverges with this change, as MySQL 8 was released on April 19, 2018. This might be useful to note as well.

Ah, that makes perfect sense. Thank you.

Is throwing an error in the JavaScript really necessary? Regardless, this works as advertised.

Weird. I cannot reproduce on Debian Trixie (13), with PHP 8.4.6 and git 2.47.2.

I need to test this still, but this looks good. Two nits, and one issue I spotted:

In D25955#26898, @valerio.bozzolan wrote:

This looks very useful. I think somewhere there is a thing that truncates this kind of pathnames to say <phorge>/path/to. I would like to invest some time to reuse that, so, no need of substr(strlen(substr(realpath

In T16035#21876, @avivey wrote:

FontAwesome doing the freemium thing makes me sad, and I'm worried about misinterpreting their license (e.g., I think the fa-regular is not part of free, but is available on the GH repo for paying customers.)

Works for me: (on rPdfe8539c6a)

I can reproduce with the embedded copy of FontAwesome. I cannot reproduce if I use Fork Awesome 1.2. Switching over as a stopgap is a potential solution, especially because it offers SVG support, which can pave the way for a native-SVG replacement.

Rebase and re-run tests with expected LANG

Update quickstart with paste mailkey removal

Capitalize consistently

In D25900#25554, @aklapper wrote:

What is (or rather was, I guess) that mailKey thingie good for?

This is great.

Much better, two nitpicks:

Setup phorge, saw nothing unusual in the setup issues.

Update documentation and simplify extension check

Ran bin/celerity map

Phorge neither uses PHP's mysqli extension nor calls PHP's htmlentities() function.

Are you sure about that? Phorge supports either the old mysql extension or mysqli - and mysql was deprecated in PHP 5.5, and removed in PHP 7.0. Notably, Phorge checks for the existence of mysqli in PhabricatorDatabaseRef, PhabricatorExtensionSetupCheck and PhabricatorPHPConfigSetupCheck.

In D25881#24406, @Cigaryno wrote:

In D25881#24380, @aklapper wrote:

Thanks. I'm personally fine with bug fixes for prototype apps.

Me too. Maybe this section should be removed?

Did some digging using the Network monitor. When using the dialog, Phorge sends this:

__csrf__: B@jqccc7335f3e341bab747f2a
__form__: 1
__dialog__: 1
tokenPHID: PHID-TOKN-emoji-7
__wflow__: true
__ajax__: true
__metablock__: 4

but using the direct URL, it sends this:

__csrf__: B@2rdakjrtaeed9a29351d9a9c
__form__: 1
__dialog__: 1

Unfortunately, all I can gleam is that the JavaScript doesn't trigger correctly when clicking on a token, and never sends the tokenPHID key-value pair as a result. It's clearly not handling the separate form correctly, because disabling JavaScript results in a proper experience.

I haven't the foggiest how this transaction is supposed to use its stub variant; attempting to create a document at /w/parent/child causes an error about /w/parent/ being required.

This doesn't seem correct. renderObject doesn't take any arguments. I suspect the intention was renderHandle, as that method takes a PHID, which the metadata key mentions.

In D25868#23446, @slip wrote:

I found this, which suggests that this shouldn't be used for headers. https://secure.phabricator.com/D21862

Followed test plan, confirmed findings, accepted revision.

In D25859#24228, @aklapper wrote:

Yes, I'd like to see this merged, [...]

Use !empty instead of isset to match previous type juggling behaviour

Remove the mailKey column from PhabricatorPaste

In D25900#24197, @valerio.bozzolan wrote:

You know better than me the context. Is there a possibility that an installation is already without that column?

Created D25900 for the deleted paste mailkey migration.

Ran ./bin/storage upgrade --namespace phabricator_quickstart_test2, and that didn't fail.

This is much better. A few spelling errors remain, plus two suggestions to make sentences read a little less awkward.

Yes, please!

In D25859#23607, @aklapper wrote:

I took the liberty to edit the commit message a bit to make reproduction easier (handy URIs to test, plus include the error message which allows easier search).

I find the current iteration too difficult to read, although that may be partially caused by the mutual language barrier.

Took some fiddling with VS2022 to find exactly what I needed to install, but worked like a charm afterwards.

Run unit tests