gmake-4.2.90 still crashes on $(wildcard hello*/)

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

gmake-4.2.90 still crashes on $(wildcard hello*/)

Gnu - Make - Bugs mailing list
$ echo 'all: ; @echo $(wildcard hello*/)' > /tmp/1.mk
$ ./make -f /tmp/1.mk
Segmentation Fault

A patch was provided here
https://lists.gnu.org/archive/html/bug-make/2017-12/msg00006.html

Any chance to have this applied for 4.3?

regards, Dmitry

_______________________________________________
Bug-make mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/bug-make
Reply | Threaded
Open this post in threaded view
|

Re: gmake-4.2.90 still crashes on $(wildcard hello*/)

Paul Smith-20
On Sun, 2019-09-01 at 16:01 -0400, Dmitry Goncharov via Bug reports and
discussion for GNU make wrote:
> $ echo 'all: ; @echo $(wildcard hello*/)' > /tmp/1.mk
> $ ./make -f /tmp/1.mk
> Segmentation Fault
>
> A patch was provided here
> https://lists.gnu.org/archive/html/bug-make/2017-12/msg00006.html
>
>
> Any chance to have this applied for 4.3?

I will apply (something close to) this patch, since it's no big deal,
but I don't understand why it's needed.

According to the POSIX spec, the gl_offs field is never used unless the
GLOB_DOOFFS flag is provided to the glob() function, which GNU make
never does.

And looking at the glob.c version we provide as part of GNU make, the
access to this field is always protected by a check for GLOB_DOOFFS as
far as I can see.  Also if I force the make version of glob.c to be
used in my testing I don't get any failures.

The only way I can see this being a problem is if (a) I misread
something in make's glob.c, or (b) there's some local version of GNU
libc glob installed on one of your systems that has a bug related to
gl_offs usage, and GNU make is choosing that instead.


_______________________________________________
Bug-make mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/bug-make