Re: [MACTCL] MacOS Debugging setup

[Phillip B. <phi...@um...>]

Can anyone try the short program below on older versions of Xcode and
MacOS?   It would be interesting to see when the issue with lldb hiding the
setting of DYLD_IMAGE_SUFFIX started.

Also, does everyone see the same result that I see with MacOS 26 and the
latest Xcode?

To run the test, cut and paste the following into a file named runit.zsh:


### cut here ###

cat > hello.c <<!

#include <stdio.h>

#include <stdlib.h>

#include <unistd.h>


int main() {

    char* var_value = getenv( "DYLD_IMAGE_SUFFIX" );

    if ( var_value ) {

        printf( "\nDYLD_IMAGE_SUFFIX is set to: %s\n", var_value );

    } else {

        printf( "\nDYLD_IMAGE_SUFFIX is not set\n\n" );

    }

}

!


export DYLD_IMAGE_SUFFIX=_debug

cc -g hello.c

echo "\nRunning without lldb:"

./a.out


echo "\nRunning with lldb:"

lldb --batch -o run ./a.out


echo ""
### cut here ###

then run it with:

$ zsh runit.zsh

Of interest is whether it prints:

DYLD_IMAGE_SUFFIX is set to: _debug


for both executions of a.out, or just for the first one, printing:


DYLD_IMAGE_SUFFIX is not set


when run under lldb.







Phil

On Fri, Oct 24, 2025 at 9:01 PM Phillip Brooks <phi...@um...> wrote:

> More digging - this article seems to bear something of a partial
> explanation:
>
>
> https://stackoverflow.com/questions/33587130/why-doesnt-lldb-forward-my-environment-variable-anymore
>
> Though the article is about 10 years old.  The article mentions that this
> restriction only occurs under certain circumstances, but I can't find any
> that it doesn't happen for.  I finally got down to this short program.  It
> is running as me, on my home directory (not a network share, not a mounted
> external volume, it's just my "Macintosh HD" home directory just as
> originally set up by Apple:
>
> % cat hello.c
>
> #include <stdio.h>
>
> #include <stdlib.h>
>
> #include <unistd.h>
>
>
> int main() {
>
>     int mypid = getpid();
>
>     printf( "PID=%d\n", mypid );
>
>     char* var_value = getenv( "DYLD_IMAGE_SUFFIX" );
>
>     if ( var_value ) {
>
>         printf( "DYLD_IMAGE_SUFFIX is set to: %s\n", var_value );
>
>     } else {
>
>         printf( "DYLD_IMAGE_SUFFIX is not set\n" );
>
>     }
>
> }
>
>
> pbrooks@Drakblar ~ % cc -g hello.c
>
> pbrooks@Drakblar ~ % ./a.out
>
> PID=37336
>
> DYLD_IMAGE_SUFFIX is set to: _debug
>
> pbrooks@Drakblar ~ % lldb ./a.out
>
> (lldb) target create "./a.out"
>
> Current executable set to '/Users/pbrooks/a.out' (arm64).
>
> (lldb) run
>
> Process 37339 launched: '/Users/pbrooks/a.out' (arm64)
>
> PID=37339
>
> DYLD_IMAGE_SUFFIX is not set
>
> Process 37339 exited with status = 0 (0x00000000)
>
> (lldb)
>
> So I can't find a case where lldb does not hide DYLD_IMAGE_SUFFIX.  It
> seems like we should have a "debug" build of tclsh that links to Tcl_debug
> directly rather than relying on this environment variable.
>
>
>
>
>
>
>