winpr_backtrace_symbols_fd ignores partial write() returns — intentional or worth fixing?

[utkonos]

Building FreeRDP 3.26.0 on Ubuntu 26.04 (gcc 15.2.0, glibc 2.43), I noticed this warning in winpr/libwinpr/utils/debug.c:

debug.c:163:31: warning: ignoring return value of 'write' declared with attribute 'warn_unused_result' [-Wunused-result]
  163 |     (void)_write(fd, lines[i], (unsigned)strnlen(lines[i], LINE_LENGTH_MAX));

FreeRDP/winpr/libwinpr/utils/debug.c

Line 163 in 1d93a2d

(void)_write(fd, lines[i], (unsigned)strnlen(lines[i], LINE_LENGTH_MAX));

The (void) cast is the conventional way to silence warn_unused_result, but gcc doesn't honor it for this attribute, so the warning fires anyway.

Looking at the code itself, there's a separate concern beyond the warning: write() is allowed to return fewer bytes than requested (short writes), and the current loop doesn't handle that case. If a short write occurs on lines[i], the remainder of that line is dropped and the loop moves on to lines[i+1], potentially producing corrupted backtrace output.

A loop that handles partial writes would both fix the latent issue and silence the warning as a side effect. Something like:

for (size_t i = 0; i < used; i++) {
    const char *p = lines[i];
    size_t remaining = strnlen(lines[i], LINE_LENGTH_MAX);
    while (remaining > 0) {
        ssize_t n = _write(fd, p, remaining);
        if (n < 0) {
            if (errno == EINTR) continue;
            break;  /* unrecoverable; give up on this line */
        }
        p += n;
        remaining -= (size_t)n;
    }
}

Not blocking my build — just flagging in case it's of interest.

[akallabeth]

@utkonos well, the function is currently unused so it did not get much attention.
long story short, pull requests always welcome.

[utkonos]

Thanks for the quick response. I will work up a PR. I didn't want to open an issue or a PR without checking first.