New libc-test regression test for uselocale [Was: Re: [musl] musl 1.1.10 released]

[Rich Felker <dalias@libc.org>]

[-- Attachment #1: Type: text/plain, Size: 1967 bytes --]

On Fri, Jun 05, 2015 at 11:01:56AM -0400, Rich Felker wrote:
> On Fri, Jun 05, 2015 at 11:20:30AM +0300, Timo Teras wrote:
> > On Thu, 4 Jun 2015 16:30:18 -0400
> > Rich Felker <dalias@libc.org> wrote:
> > 
> > > This release fixes regressions introduced as part of the dynamic
> > > linker bootstrap overhaul in musl 1.1.9, and adds several new features
> > > and improvements. Internal cleanup and optimizations have been made to
> > > the locale system. A fail-safe/allocation-free locale_t object for the
> > > C locale is now available via newlocale(), and the iconv_open function
> > > now supports "" and "CHAR" as aliases for the native (UTF-8) encoding.
> > > A new crt start file, rcrt1.o, is provided for producing static-linked
> > > position independent executables (PIE). Minor PIE-related
> > > arch-specific bugs, and a bug in the ungetc and ungetwc stdio
> > > functions which caused them to fail on files in EOF status, have also
> > > been fixed. 
> > > 
> > > http://www.musl-libc.org/releases/musl-1.1.10.tar.gz
> > > http://www.musl-libc.org/releases/musl-1.1.10.tar.gz.asc
> > > 
> > > Thanks as always to musl's Patreon release sponsors:
> > > 
> > > - The Midipix Project (midipix.org)
> > > - Hurricane Labs (hurricanelabs.com)
> > > - Justin Cormack
> > > 
> > > Further notes on upcoming development will follow soon.
> > 
> > There's relatively nasty regression in uselocale() breaking most X
> > applications (due to some libraries).
> > 
> > See patch at:
> > http://git.alpinelinux.org/cgit/aports/plain/main/musl/0001-fix-uselocale-locale_t-0-to-not-modify-locale.patch
> 
> Uhg, this kind of thing is exactly why I've been wishing we had tests
> for locale functions. I tested it minimally by hand but didn't think
> to try passing (locale_t)0. I'll probably make a fix release with this
> patch applied.

Regression test for just this one issue is attached. We should still
have some big locale functionality tests, though...

Rich

[-- Attachment #2: uselocale-0.c --]
[-- Type: text/plain, Size: 607 bytes --]

// commit: 63f4b9f18f3674124d8bcb119739fec85e6da005
// uselocale(0) should not change the current locale
#include <locale.h>
#include "test.h"

int main(void)
{
	locale_t c = newlocale(LC_ALL_MASK, "C", 0);

	if (!c) {
		t_error("newlocale failed\n");
		return t_status;
	}

	if (!uselocale(c))
		t_error("uselocale(c) failed\n");

	locale_t l1 = uselocale(0);
	if (l1 != c)
		t_error("uselocale failed to set locale: "
			"%p != %p\n", (void*)l1, (void*)c);

	locale_t l2 = uselocale(0);
	if (l2 != l1)
		t_error("uselocale(0) changed locale: "
			"%p != %p\n", (void*)l2, (void*)l1);

	return t_status;
}