I've submitted this bug more than 1 year ago at
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=163306. Now I can supply
new information: I just confirmed that the bug still exists in latest official
release version 4.1.1 as well as in a gcc trunk build from today "gcc (GCC)
4.2.0 20060729 (experimental)". Had I realised last year that this is a problem
in the compiler per se and not in any GNU/Linux distribution, I would have
probably entered the bug here in the first place.
I copy / paste here from the other URL. Note that there are existing comments
there.
How reproducible:
Always
Steps to Reproduce:
1. Compile the program
#include <stdio.h>
#include <netdb.h>
int main()
{
printf("%d", h_errno);
return 0;
}
and the program
#include <netdb.h>
int main()
{
h_errno = 0;
return 0;
}
with mudflap enabled, i.e.
gcc -fmudflap prog1.c -lmudflap -o prog1
and
gcc -fmudflap prog2.c -lmudflap -o prog2
2. Run prog1 and prog2.
Actual Results: When prog1 is run, this is what I get:
*******
mudflap violation 1 (check/read): time=1121380261.574621 ptr=0xb7eff6b4 size=4
pc=0xb7f09322 location=`a.cpp:6 (main)'
/usr/lib/libmudflap.so.0(__mf_check+0x44) [0xb7f09322]
./a.out(main+0x81) [0x8048839]
/usr/lib/libmudflap.so.0(__wrap_main+0x1d8) [0xb7f0a04e]
Nearby object 1: checked region begins 17B after and ends 20B after
mudflap object 0x80cb430: name=`errno area'
bounds=[0xb7eff6a0,0xb7eff6a3] size=4 area=static check=0r/0w liveness=0
alloc time=1121380261.574591 pc=0xb7f09e0a
number of nearby objects: 1
When prog2 is run, the report is similar, but it says "(check/write)", instead
of "(check/read)" -- because h_errno is written to.
Expected Results: There should have been no violation messages.
--
Summary: Reading or assigning global h_errno variable causes a
memory violation report
Product: gcc
Version: 4.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libmudflap
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: vesselinpeev at hotmail dot com
GCC host triplet: UNCO Running large program compiled with
mudflap abort
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28536
