Fix GH-13932: Attempt to fix mbstring on windows build (msvc). - #13933

These can get out of sync, use a define instead.

I see

Thanks. This should go into branch PHP-8.3. And then @bukka can cherry pick this into PHP-8.3.5

oh gotcha was not sure about this. cheers.

Merged via 2cfd9df1

it cannot go to 8.3.5 as it is already tagged. We might need to do immediate 8.3.6 instead.

@bukka I thought standard procedure were a retag for such cases? As long as it's not announced, that is. (At least retags did happen in the past.)

@bukka I thought standard procedure were a retag for such cases? As long as it's not announced, that is. (At least retags did happen in the past.)

We no longer retag for good reasons: #11868

@nielsdos Ah, I just wanted to bring 8.2.9 as the example where we retagged in the past.
I do however think retagging is still the right way. People should not assume tag == release. But anyway, it seems policy now then? Then disregard my comments. I don't want to start a discussion here.

@nielsdos Ah, I just wanted to bring 8.2.9 as the example where we retagged in the past. I do however think retagging is still the right way. People should not assume tag == release.

The additional problem however is how git handles force pushes to tags (see the issue).

@nielsdos While we are here, can you please remind me the reason why we have two different days for tagging and releasing? Can't the builds be run just on the HEAD of the php-x.y.z branch, triggered manually once ready? Rather than a tag.

@nielsdos While we are here, can you please remind me the reason why we have two different days for tagging and releasing? Can't the builds be run just on the HEAD of the php-x.y.z branch, triggered manually once ready? Rather than a tag.

I have no idea honestly, I can only imagine it is to do last minute checks if something wrong went into the tag, but that's speculation.

No re-tagging is just completely wrong approach that goes against all good practices. So it's not going to be done in that way.

I think we should remove this delay of two days as well as it's a big flaw in our security disclosure handling.

I would suggest to never use VLA again :) regardless of compiler support, rarely needed in practice I think.