Re: [PATCH] libio: Fix open_memstream flush (NULL)

[Florian Weimer <fw at deneb dot enyo dot de>]

* Adhemerval Zanella:

> POSIX specifies that the state and size after a fflush on a stream
> opened by open_memstream should be updated and current implementation
> does not act accordingly on a fflush (NULL) (meant to act on all opened
> streams).
>
> This patch fixes it for open_{w}memstream and also adjust the
> tst-memstream3 to use libsupport.

Subject is wrong, it should be fflush.

Please put the test case into a new file, and change it to call fflush
(NULL); it currently calls fflush (fp).  As a result, it does not
actually test the change.

You also need to verify that this does not call malloc:

#include <stdio.h>
#include <stdlib.h>
#include <err.h>

int
main (void)
{
  char * buffer = NULL;
  size_t length = 0;
  FILE *fp = open_memstream (&buffer, &length);
  if (fp == NULL)
    err (1, "open_memstream");
  abort ();
}

I think we don't want the malloc call for a call to exit, either.
Similarly for __stack_chk_fail, but that's something we need to fix
there.

> +static int
> +_IO_mem_overflow (_IO_FILE *fp, int c)
> +{
> +  int ret = _IO_str_overflow (fp, c);
> +
> +  struct _IO_FILE_memstream *mp = (struct _IO_FILE_memstream *) fp;
> +  *mp->sizeloc = fp->_IO_write_ptr - fp->_IO_write_base;
> +
> +  return ret;
> +}

_IO_str_overflow does not correctly process an EOF argument, but
_IO_flush_all_lockp calls the overflow function with EOF.  This is the
reason why the test with fflush (NULL) fails.

The most direct fix appears to be to call _IO_mem_sync for c == EOF.

Let me suggest again that libio changes are not appropriate during the
freeze.