[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <52DE631F-9A10-42E5-A72B-9CD282EB61CB@inria.fr>
Date: Mon, 03 Jul 2023 23:23:00 +0200
From: Jens Gustedt <jens.gustedt@...ia.fr>
To: musl@...ts.openwall.com
Subject: Re: fix various warnings/theoretical UB
Hello,
Am 3. Juli 2023 21:59:57 MESZ schrieb Rich Felker <dalias@...c.org>:
> On Mon, Jul 03, 2023 at 01:55:57PM -0400, Alex Xu (Hello71) wrote:
> > See attached patches.
>
> > From 978f2cded65ce73450277d3fde48f038b339d5f9 Mon Sep 17 00:00:00 2001
> > From: "Alex Xu (Hello71)" <alex_y_xu@...oo.ca>
> > Date: Sun, 2 Jul 2023 20:28:23 -0400
> > Subject: [PATCH 1/4] volatile static -> static volatile
> >
> > C11 6.11.5p1:
> >
> > > The placement of a storage-class specifier other than at the
> > > beginning of the declaration specifiers in a declaration is an
> > > obsolescent feature.
> >
> > gcc also warns about this.
> > ---
> > src/time/timer_create.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/src/time/timer_create.c b/src/time/timer_create.c
> > index cd32c945..9216b3ab 100644
> > --- a/src/time/timer_create.c
> > +++ b/src/time/timer_create.c
> > @@ -61,7 +61,7 @@ static void *start(void *arg)
> >
> > int timer_create(clockid_t clk, struct sigevent *restrict evp, timer_t *restrict res)
> > {
> > - volatile static int init = 0;
> > + static volatile int init = 0;
> > pthread_t td;
> > pthread_attr_t attr;
> > int r;
> > --
> > 2.41.0
>
> No objection to this change. It's contrary to usual style. I would say
> let's convert to pthread_once, but this code is slated for removal
> anyway once signals are no longer used for SIGEV_THREAD timers.
>
> > From b98f243e7921ddff6978ee9b0ce9f08efaa17951 Mon Sep 17 00:00:00 2001
> > From: "Alex Xu (Hello71)" <alex_y_xu@...oo.ca>
> > Date: Sun, 2 Jul 2023 20:29:41 -0400
> > Subject: [PATCH 2/4] __year_to_secs: fix dangling pointer
> >
> > C11 6.5.2.5p5:
> >
> > > If the compound literal occurs outside the body of a function, the
> > > object has static storage duration; otherwise, it has automatic
> > > storage duration associated with the enclosing block.
> >
> > gcc also warns about this.
> > ---
> > src/time/__year_to_secs.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/src/time/__year_to_secs.c b/src/time/__year_to_secs.c
> > index 2824ec6d..d215880a 100644
> > --- a/src/time/__year_to_secs.c
> > +++ b/src/time/__year_to_secs.c
> > @@ -10,9 +10,9 @@ long long __year_to_secs(long long year, int *is_leap)
> > return 31536000*(y-70) + 86400*leaps;
> > }
> >
> > - int cycles, centuries, leaps, rem;
> > + int cycles, centuries, leaps, rem, tmp;
> >
> > - if (!is_leap) is_leap = &(int){0};
> > + if (!is_leap) is_leap = &tmp;
> > cycles = (year-100) / 400;
> > rem = (year-100) % 400;
> > if (rem < 0) {
> > --
> > 2.41.0
>
> Seems like a bogus warning. The enclosing block is the whole function,
No, the `if` statement forms a block of itself, and then the dependent statement forms yet another block.
We rectify the terminology a bit in C23 hopefully make it easier to read without changing semantics
> the same as the lifetime of the pointer. This might merit
> investigation on whether GCC is doing something wrong though..
>
> > From a30c4ab397af040d10d978d97dd4a6835d4b99a8 Mon Sep 17 00:00:00 2001
> > From: "Alex Xu (Hello71)" <alex_y_xu@...oo.ca>
> > Date: Sun, 2 Jul 2023 20:54:45 -0400
> > Subject: [PATCH 3/4] fix mismatched VLA parameter types
> >
> > gcc warns about this, and it's probably technically UB
> > ---
> > src/internal/procfdname.c | 2 +-
> > src/prng/seed48.c | 2 +-
> > 2 files changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/src/internal/procfdname.c b/src/internal/procfdname.c
> > index fd7306ab..bfa3e7e5 100644
> > --- a/src/internal/procfdname.c
> > +++ b/src/internal/procfdname.c
> > @@ -1,6 +1,6 @@
> > #include "syscall.h"
> >
> > -void __procfdname(char *buf, unsigned fd)
> > +void __procfdname(char buf[static 15+3*sizeof(int)], unsigned fd)
> > {
> > unsigned i, j;
> > for (i=0; (buf[i] = "/proc/self/fd/"[i]); i++);
>
> This was raised/proposed before and is probably an okay change, but
> I'd like to understand what the reason "it's probably technically UB"
> is.
>
> > diff --git a/src/prng/seed48.c b/src/prng/seed48.c
> > index bce7b339..7b789086 100644
> > --- a/src/prng/seed48.c
> > +++ b/src/prng/seed48.c
> > @@ -2,7 +2,7 @@
> > #include <string.h>
> > #include "rand48.h"
> >
> > -unsigned short *seed48(unsigned short *s)
> > +unsigned short *seed48(unsigned short s[3])
> > {
> > static unsigned short p[3];
> > memcpy(p, __seed48, sizeof p);
> > --
>
> This one is almost surely not UB because there's no static and the 3
> is ignored. The question is just whether the static produces a
> difference in the declaration type that makes them clash.
>
> Rich
Jens
--
Jens Gustedt - INRIA & ICube, Strasbourg, France
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
