I have a closed source C library my Linux application depends on.
This library isn’t particularly well designed, and most of the functions in it print a bunch of junk to stdout and stderr as a side effect. Sometimes this is useful for debugging, but most of the time I’d like to discard it.
Is there any way for me to “capture” stdout/stderr on those function calls only, while keeping it available for the main program?
In C I’ve used a strategy with dup/dup2 before, something like this where the file descriptors are temporariliy redirected somewhere else.
In principle you could wrap the library calls so that you dup2() before the syscall and then again after the call to restore the original file descriptor…
You could also try to mix running the library on a separate process, with using dup2 to change where stdout / stderr go for that process. Hard to say if this is doable in your case without more details.
This is in fact what I ended up doing. I don’t like it, but I don’t think there’s a better way. If the code I’m interfacing with was well-made I wouldn’t need it in the first place, but such is life sometimes when you’re dealing with proprietary legacy bullshit.
At least Zig’s defer keyword makes it relatively clean. I have a redirect(output fd) and a restore(output fd) function that perform the operations with dup2.
Call the redirect function at the beginning of a scope, defer the restore, then continue as usual.
