You can check if the optimizelegibility is indeed the cause by disabling it via the web inspector. It is the issue for me, at least.

This is due to a bug in Chrome (most likely) which makes text-rendering: optimizeLegibility behave very strangely.

Qafoolabs mentions a "proposed timing API" in a pull request. I can't seem to find anything about anything having been proposed, but it might be worthwhile to keep that in mind.

I do see that gc_runs and collected are available (via GC_G(...)) unconditionally? It's not much, but it's more than is available now.

Pretty sure you've been across the same path as I was, but I want to share my findings just in case someone can find a way to use it:

Ah, no - I was wrong. It's because of the ---- line from the quoted section (- indicates a list item too). Observe:

> --
> Hello
> 1. One
> 2. Two

Hello

1. One
2. Two

I'm guessing that this is a side-effect from rPHU0db4b149c4e0. Strange though. :/

Specific link to the source, for future reference: http://git.php.net/?p=php-src.git;a=blob;f=ext/standard/exec.c;h=6106fe2c86bb687979e36e2070a99283cc0a0878;hb=HEAD#l365

See T6348 for details on why this happens. escapeshellarg under Windows removed % signs.

I believe the check for TERM should be fine. However, for reference, the Symfony framework uses a slightly simpler check - and will probably behave saner in most of these cases to do go wrong.

Now with lint and unit tests running

Basically, double quotes disappear when passed through escapeshellarg. The command for history basically becomes:

The history support is useful. I might be able to find out what's happening.

Oh, I notice I forgot to mention it was about a task :/ My apologies for
that.

I've now seen the same issue over several commits, leading me to believe
that it's not a fluke.

It's not very rare, it's just that MySQL (by default) will allow non-standard behavior for the GROUP BY clause.

A related case with try..catch..finally (see D10337) works in the same way as @epriestley described. However, in that case it's easy to fix. This may be less trivial, but still worth fixing.

Having the favicon also change would be very cool. I usually have my phabricator tab pinned, which means that I don't see the title.

• However, I'm not sure if there's an agreed-upon or standard way to canonicalize combining marks.

Note that the token constant values have changed because I built with Bison 3.0.2 instead of 2.7

@epriestley: could you take a look at the comment above? I'd push out a diff but for some reason the values of the tokens change when I rebuild xhpast.

Verified to work, thanks!

Catches are optional when you have a while. This small diff fixes that:

Thanks!

Is this warning bugged because I'm using braces in the case statement? It doesn't occur for the other fallthrough statement at the end of the build step.

The braces aren't necessary and probably cause the message.

Debug-less

You could presumably also use identify from ImageMagick. Probably overkill though.

This could of course break if people depend on arc feature with no branches/bookmarks having exit code 1.

Well, I seem to remember it wrong; when I close the new-commit file in Sublime Text now, arc continues as it should.

If I recall correctly, I had the same kind-of issue. The file opened in an already opened instance, and arc would only continue if I closed the entire instance. Adding -n helped in that only the new instance had to be closed.

Use a simpler grammatical construct for the error message

I don't know, it was pretty late and I just pushed it out hoping for feedback. :-)

Closed by commit rARC3201239e85ea.

This probably needs quite some work.

It's crazy, but it does work.

Taken on Windows running cmd.exe:

It seems that this breaks when the user doesn't have MingW installed (which I don't have anymore - just did a reinstall).

Not an actual update, just running with sublime text as well to be sure it's handled correctly.

It's Windows, not that great a surprise actually...

This now suddenly runs cleanly on my machine. Probably thanks to the reboot I just did. :/

Separate out the template parameter

This time I was smart enough to test it with a merge commit. The result looked okay to me.

Don't use the actual newline since Windows doesn't support it.

Windows is so awesome it doesn't need newlines. Oh wait, it sucks!

That should be handled by csprintf, right?

One step I actually did differently was running mingw32-make in my regular cmd. You only have to follow the steps here and it works just fine.

Closed by commit rPHU3fe4daa19d2c.

Remove the rebuild xhpast.exe and expand the comment a bit.

The steps pretty much seem the same as what I have experienced.

This very much seems okay to me.

The fix @epriestley mentioned, indeed.

I can confirm that your fix makes the problem go away at my Windows pc. Nice :)

The hg grep search doesn't seem to be too slow. Also, by replacing it with a hg locate command you lose the line numbers which seems to be a shame.