`StackTrace.format` is not thread-safe due to the usage of `getenv` in tty detect

[jackalcooper]

in tty.zig

 pub fn detect(file: File) Config {
 const force_color: ?bool = if (builtin.os.tag == .wasi)
 null // wasi does not support environment variables
 else if (process.hasNonEmptyEnvVarConstant("NO_COLOR"))
 false
 else if (process.hasNonEmptyEnvVarConstant("CLICOLOR_FORCE"))
 true
 else
 null;

process.hasNonEmptyEnvVarConstant calls posix.getenv(key) which is not thread-safe on glibc linux

gdb

Thread 6 "2_scheduler" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffb3b94640 (LWP 3483)]
0x00007fffaba5032e in posix.getenv (key=...) from /home/runner/work/kinda/kinda/kinda_example/_build/test/lib/kinda_example/priv/lib/libKindaExampleNIF.so
#0 0x00007fffaba5032e in posix.getenv (key=...)
 from /home/runner/work/kinda/kinda/kinda_example/_build/test/lib/kinda_example/priv/lib/libKindaExampleNIF.so
#1 0x00007fffaba63e8b in process.hasNonEmptyEnvVarConstant__anon_11739 (
 key=...)
 from /home/runner/work/kinda/kinda/kinda_example/_build/test/lib/kinda_example/priv/lib/libKindaExampleNIF.so
#2 0x00007fffaba340a4 in Io.tty.Config.detect (file=...)
 from /home/runner/work/kinda/kinda/kinda_example/_build/test/lib/kinda_example/priv/lib/libKindaExampleNIF.so
#3 0x00007fffaba0e2e8 in Io.tty.detectConfig (file=...)
 from /home/runner/work/kinda/kinda/kinda_example/_build/test/lib/kinda_example/priv/lib/libKindaExampleNIF.so
#4 0x00007fffabb0e927 in builtin.StackTrace.format (self=..., 
 writer=0x7fffb3b8f5d0)

[mlugg]

It also, of course, doesn't make sense for this function to be requesting the ttyconfig for stderr anyway; there's a TODO comment in the implementation of StackTrace.format about this.

I think the default format method for StackTrace should be using .no_color, and there should be an alternative formatter (fmtColor?) which takes in a std.Io.tty.Config.

[GasInfinity]

I think this can be closed as #25592 moved the formatter to std.debug.FormatStackTrace

[KilianHanich]

Tbf, it is only a problem if you also use setenv with glibc (and on macOS you get memory leaks from calling it), which is probably a bad idea anyway if you want to have your code be portable since some high security operating systems have environment variables as read-only (but then again, porting Zig's standard library to some of these is often impossible anyway; try to port for example std.process.Child's API to Fuchsia, you would need API changes).