7.4 fpm causes segfaults anytime gettext is invoked, M2 Apple Silicon

[Saeven]

Hey everyone.

Been banging my head on this one a bit.

Am on an M2 Mac, and this test script will cause a crash if accessed via php-fpm, but runs fine php cli. Has anyone run into this?

<?php

echo gettext("Hello there");

Running /opt/homebrew/opt/[email protected]/sbin/php-fpm --nodaemonize while doing this yields:

[24-Jul-2023 14:17:12] WARNING: [pool www] child 10691 exited on signal 11 (SIGSEGV) after 3.445288 seconds from start

Conversely, changing that script to a simple:

<?php

echo "Hello there";

or

<?php

phpinfo();

Works as expected.

Thanks!

[jstrellner]

I am running into the same issue. I've tried it against [email protected] and [email protected] and it's the same. I'm running an M3 Max.

echo gettext("hello");

This will always trigger it.

Putting putenv("LC_ALL=C"); above the echo line above will make it work, however that doesn't solve my actual main issue.

My main issue is that installing mongodb makes php stop working entirely with:
[Tue Jan 16 15:29:54.667430 2024] [core:notice] [pid 59504] AH00052: child pid 59509 exit signal Segmentation fault (11)

I need the mongo driver for a project I'm working on, but when installed apache/php doesn't work anywhere, even scripts that don't call mongo. I think the mongo error and the gettext error are related.

Any ideas on how to fix this? I'm on my second day trying to fix this.

[Saeven]

Well, looks like this is back. I'm on Sequoia 15.4, this repo's PHP 7.4, and the old modification that used to fix this (of the home-brew launch file):

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
        <key>EnvironmentVariables</key>
        <dict>
            <key>OBJC_DISABLE_INITIALIZE_FORK_SAFETY</key>
            <string>YES</string>
        </dict>
        <key>KeepAlive</key>

No longer seems to do the trick.

Adding putenv("LC_ALL=C"); does seem to band-aid things for now.

[jacobsenj]

Also happens on M1 Apple Silicon. The putenv("LC_ALL=C"); is confirmed working.