Hacker News Reader

Comment by AaronBallman | Hacker News Reader

Funny you should mention that, as that feature has come up recently in mailing list discussions. We have not seen an actual proposal for adopting it yet, but features similar semantics are being discussed as a possible idea (no promises).

FWIW, I don't think it would wind up being spelled with attribute syntax because we would likely want programmers to have a guarantee that the cleanup will happen (and attributes can be ignored by the implementation).

View on HN

Hopefully it'd at least be syntactically similar, so we can have an

  #ifdef __STDC_CLEANUP__
  #define my_cleanup(func) stdc_cleanup(func)
  #else
  #define my_cleanup(func) __attribute__((cleanup(func)))
  #endif

i.e. it would require that it at least goes in the same places as an attribute.

I believe the last proposal was in 2008 (ignore the try..finally stuff here): http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1298.pdf

So I guess it needs someone to take that and update it, also to pull up a full list of current Linux software which is using this feature (which as I say these days is a surprising amount).

Here's our usage: https://github.com/FRRouting/frr/blob/master/lib/privs.h#L14...

  #define frr_with_privs(privs)                                                  \
          for (struct zebra_privs_t *_once = NULL,                               \
                                    *_privs __attribute__(                       \
                                            (unused, cleanup(_zprivs_lower))) =  \
                                            _zprivs_raise(privs, __func__);      \
               _once == NULL; _once = (void *)1)

This gives us a block construct that guarantees elevated privileges are dropped when the block is done:

  frr_with_privs(privs) {
    ... whatever ...
    break;  /* exit block, drop privileges */
    return; /* return, drop privileges */
  }

We have a nice macro for acquiring locks that only applies to the scope:

https://github.com/libguestfs/nbdkit/blob/e58d28d65bfea3af36...

You end up with code like this:

https://github.com/libguestfs/nbdkit/blob/e58d28d65bfea3af36...

It's so useful to be able to be sure the lock is released on all return paths. Also because it's scope-level you can scope your locks tightly to where they are needed.

We use it extensively in our proprietary codebases as well, FWIW. Not real open data for me to point to, but: a few million lines of C, and a handful of billion USD in revenue. If that helps weigh in on "yes, please standardize this common practice."

Reply on news.ycombinator.com