[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 Jeffrey A. Law changed: What|Removed |Added Target Milestone|8.0 |9.0 --- Comment #13 from Jeffrey A. Law --- I think this needs to defer to gcc-9.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 --- Comment #12 from H.J. Lu --- (In reply to Martin Sebor from comment #11) > > Until then, the only way to avoid the warning that I can think of is to > exclude from checking arguments passed through the ellipsis, but that will > result in some false negatives for sprintf et al. So the question is: do > the benefits of the warning outweigh the cost of the false positives. > It is a regression that GCC 8 can't compile lib/str_error_r.c in Linux kernel without modifications.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919
Martin Sebor changed:
What|Removed |Added
See Also||https://gcc.gnu.org/bugzill
||a/show_bug.cgi?id=83688,
||https://gcc.gnu.org/bugzill
||a/show_bug.cgi?id=83859
--- Comment #11 from Martin Sebor ---
The warning will go away once -Wrestrict is implemented for sprintf, hopefully
in GCC 9 (the patch I submitted in July didn't make it to GCC 8:
https://gcc.gnu.org/ml/gcc-patches/2017-07/msg00036.html), and the handling for
the built-in functions removed from the front-end. This is being tracked in
bug 83688.
Until then, the only way to avoid the warning that I can think of is to exclude
from checking arguments passed through the ellipsis, but that will result in
some false negatives for sprintf et al. So the question is: do the benefits of
the warning outweigh the cost of the false positives.
The potential for false positives here also isn't limited to sprintf. It's a
general problem that affects user-defined functions as well. For example:
$ cat z.c && gcc -S -Wall z.c
void f (char* restrict, const char* restrict, const void*);
void g (char *s)
{
f (s, "", s);
}
z.c: In function ‘g’:
z.c:5:3: warning: passing argument 1 to restrict-qualified parameter aliases
with argument 3 [-Wrestrict]
f (s, "", s);
^
The only way to avoid these false positives (and negatives) is to add some
annotation to let users indicate how each argument is used (e.g., such as the
attributes read_only and write_only that I've been working on). This is being
tracked in bug 83859.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 --- Comment #10 from H.J. Lu --- (In reply to H.J. Lu from comment #9) > (In reply to Jakub Jelinek from comment #8) > > Not aware of any right now, but 1) if the format string is not compile time > > known, we don't know if it is %p or not 2) the format string parsing is done > > We only need to handle the known format string case. > > > in other passes (-Wformat code in FEs, and gimple-ssa-sprintf.c) compared to > > the restrict warning. > > This shouldn't be the reason not to do it. If there is no way to fix it, we should warn (char *) cases under -Wextra, not under -Wall.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 --- Comment #9 from H.J. Lu --- (In reply to Jakub Jelinek from comment #8) > Not aware of any right now, but 1) if the format string is not compile time > known, we don't know if it is %p or not 2) the format string parsing is done We only need to handle the known format string case. > in other passes (-Wformat code in FEs, and gimple-ssa-sprintf.c) compared to > the restrict warning. This shouldn't be the reason not to do it.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 --- Comment #8 from Jakub Jelinek --- Not aware of any right now, but 1) if the format string is not compile time known, we don't know if it is %p or not 2) the format string parsing is done in other passes (-Wformat code in FEs, and gimple-ssa-sprintf.c) compared to the restrict warning.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 Richard Biener changed: What|Removed |Added Keywords||diagnostic Target Milestone|--- |8.0
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 --- Comment #7 from H.J. Lu --- (In reply to Jakub Jelinek from comment #6) > On the other side, the warning matches the documented behavior and it might > be too difficult to add all the exceptions when we know that the argument > will not be really dereferenced. How many exceptions are there, beside %p?
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 --- Comment #6 from Jakub Jelinek --- On the other side, the warning matches the documented behavior and it might be too difficult to add all the exceptions when we know that the argument will not be really dereferenced.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 Jakub Jelinek changed: What|Removed |Added CC||jakub at gcc dot gnu.org --- Comment #5 from Jakub Jelinek --- Yeah, only arguments corresponding to %s/%ls/%S/%n or whatever else can actually dereference the pointer (for *scanf of course all the pointers).
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 --- Comment #4 from H.J. Lu --- (In reply to Jeffrey A. Law from comment #3) > Umm, isn't that invalid code? The whole point of the warning is to catch > precisely this kind of broken idiom. Right? Why is it invalid? Argument 4 is for %p. It looks valid to me.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 Jeffrey A. Law changed: What|Removed |Added CC||law at redhat dot com --- Comment #3 from Jeffrey A. Law --- Umm, isn't that invalid code? The whole point of the warning is to catch precisely this kind of broken idiom. Right?
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919 H.J. Lu changed: What|Removed |Added CC||msebor at gcc dot gnu.org --- Comment #2 from H.J. Lu --- This is caused by r258455.
[Bug c/84919] [8 Regression] error: passing argument 1 to restrict-qualified parameter aliases with argument 5 [-Werror=restrict]
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84919
H.J. Lu changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed||2018-03-16
Ever confirmed|0 |1
--- Comment #1 from H.J. Lu ---
[hjl@gnu-bdx-1 tmp]$ cat x.c
#include
#include
char *str_error_r(char *buf, size_t buflen)
{
snprintf(buf, buflen, "(%p, %zd)", buf, buflen);
return buf;
}
[hjl@gnu-bdx-1 tmp]$ /usr/gcc-8.0.1-x32/bin/gcc -O2 -S -Wall x.c
x.c: In function ‘str_error_r’:
x.c:5:2: warning: passing argument 1 to restrict-qualified parameter aliases
with argument 4 [-Wrestrict]
snprintf(buf, buflen, "(%p, %zd)", buf, buflen);
^~~~
[hjl@gnu-bdx-1 tmp]$
