On 10/13/21 01:31, Akim Demaille wrote:
> Hi Dennis,
>
>> Le 13 oct. 2021 à 00:19, Dennis Clarke <dcla...@blastwave.org> a écrit :
>>
>> On 10/9/21 04:35, Akim Demaille wrote:
>>> Hi Dennis,
>>>
>>> Thanks for this bug report.
>>
>> The least I can do.
>
> :)
>
>> You will love this :
>>
>> p = 0x100104410
>> (dbx) print *p
>> *p = {
>> name = 0x105bd0 "<bad address 0x0000000000105bd0>"
>> type = 260
>> value = {
>> var = 2.1225218246347e-314
>> fun = 0x100103f00
>> }
>> next = 0x1001043e0
>> }
>> (dbx) quit
>> beta $
>
> Well, maybe the problem is actually very simple: the sym_table pointer is not
> initialized to NULL by the compiler. And as a matter of fact, it is probably
> right, I don't believe the C standard requires it to do it. It is mandatory
> for static variables, but that's not the case here.
>
I can not think of any reason where a pointer is initialized to NULL.
If we assume otherwise then we get nose demons right ? :)
> Could you look for the following lines in examples/c/mfcalc/mfcalc.y
>
> /* The symbol table: a chain of 'struct symrec'. */
> symrec *sym_table;
>
> and turn it into this?
>
> /* The symbol table: a chain of 'struct symrec'. */
> symrec *sym_table = NULL;
>
> That should do it. Thanks!
>
Sorry, no joy.
beta $ cat ./test-suite.log
=======================================
GNU Bison 3.8.2: ./test-suite.log
=======================================
# TOTAL: 8
# PASS: 4
# SKIP: 2
# XFAIL: 0
# FAIL: 2
# XPASS: 0
# ERROR: 0
.. contents:: :depth: 2
FAIL: examples/c/mfcalc/mfcalc
==============================
checking for diff --strip-trailing-cr...
diff: illegal option -- strip-trailing-cr
usage: diff [-bitw] [-c | -e | -f | -h | -n | -u] file1 file2
diff [-bitw] [-C number | -U number] file1 file2
diff [-bitw] [-D string] file1 file2
diff [-bitw] [-c | -e | -f | -h | -n | -u] [-l] [-r] [-s] [-S
name] directory1 directory2
checking for diff --strip-trailing-cr...
mfcalc: PASS: 1
mfcalc: PASS: 2
mfcalc: PASS: 3
mfcalc: FAIL: 4 (expected status: 0, effective: 139)
./examples/test: line 50: 5337 Segmentation Fault (core dumped)
"$abs_medir/$me" "$@"
mfcalc: PASS: 5
FAIL examples/c/mfcalc/mfcalc.test (exit status: 1)
SKIP: examples/c/bistromathic/bistromathic
==========================================
checking for diff --strip-trailing-cr...
diff: illegal option -- strip-trailing-cr
usage: diff [-bitw] [-c | -e | -f | -h | -n | -u] file1 file2
diff [-bitw] [-C number | -U number] file1 file2
diff [-bitw] [-D string] file1 file2
diff [-bitw] [-c | -e | -f | -h | -n | -u] [-l] [-r] [-s] [-S
name] directory1 directory2
checking for diff --strip-trailing-cr...
./examples/test: line 50: 5448 Segmentation Fault (core dumped)
"$abs_medir/$me" "$@"
checking for readline output...
1,3d0
< > 0
< 0
< >
1d0
< 0
SKIP: this is not the GNU Readline we expect
SKIP examples/c/bistromathic/bistromathic.test (exit status: 77)
FAIL: examples/c/glr/c++-types
==============================
checking for diff --strip-trailing-cr...
diff: illegal option -- strip-trailing-cr
usage: diff [-bitw] [-c | -e | -f | -h | -n | -u] file1 file2
diff [-bitw] [-C number | -U number] file1 file2
diff [-bitw] [-D string] file1 file2
diff [-bitw] [-c | -e | -f | -h | -n | -u] [-l] [-r] [-s] [-S
name] directory1 directory2
checking for diff --strip-trailing-cr...
c++-types: FAIL: 1 (expected status: 0, effective: 139)
./examples/test: line 50: 5549 Segmentation Fault (core dumped)
"$abs_medir/$me" "$@"
FAIL examples/c/glr/c++-types.test (exit status: 1)
SKIP: examples/c/reccalc/reccalc
================================
checking for diff --strip-trailing-cr...
diff: illegal option -- strip-trailing-cr
usage: diff [-bitw] [-c | -e | -f | -h | -n | -u] file1 file2
diff [-bitw] [-C number | -U number] file1 file2
diff [-bitw] [-D string] file1 file2
diff [-bitw] [-c | -e | -f | -h | -n | -u] [-l] [-r] [-s] [-S
name] directory1 directory2
checking for diff --strip-trailing-cr...
SKIP: gimme one seq
SKIP examples/c/reccalc/reccalc.test (exit status: 77)
beta $
>From a testing perspective I get it that sometimes a segfault is what we
want. However I am guessing we don't want that today :
void example()
{
some_ptr *p=NULL;
if(expr1)
{
p=foobar();
}
else if(expr2)
{
p=barbaz();
}
do_something(p);
}
So, yeah, if neither expr1 or expr2 then kaboom.
Looking at the first kaboom we see the same situation we had before :
beta $ dbx examples/c/mfcalc/mfcalc
time_1634114818-pid_5337-uid_16411-gid_20002-fid_mfcalc.core
Reading mfcalc
core file header read successfully
Reading ld.so.1
Reading libm.so.2
Reading libc.so.1
program terminated by signal SEGV (no mapping at the fault address)
0xffffffff7eb3c18c: strcmp+0x00cc: ldx [%o1 + %o2], %o3
Current function is getsym
1461 if (strcmp (p->name, name) == 0)
(dbx) where
[1] strcmp(0x105bd0, 0x100106d60, 0xfffffffeffffee70, 0x0,
0x8080808080808080, 0x101010101010101), at 0xffffffff7eb3c18c
=>[2] getsym(name = 0x100106d60 "a"), line 1461 in "mfcalc.c"
[3] yylex(), line 1514 in "mfcalc.c"
[4] yyparse(), line 1076 in "mfcalc.c"
[5] main(argc = 2, argv = 0xffffffff7ffff278), line 1537 in "mfcalc.c"
(dbx) list
1461 if (strcmp (p->name, name) == 0)
1462 return p;
1463 return NULL;
1464 }
1465
1466 #include <ctype.h>
1467 #include <stddef.h>
1468
1469 int
1470 yylex (void)
(dbx) print &p
&p = 0xffffffff7fffe758
(dbx) print &name
&name = 0xffffffff7fffe768
(dbx) print name
name = 0x100106d60 "a"
(dbx) print p
p = 0x100104410
(dbx) print *p
*p = {
name = 0x105bd0 "<bad address 0x0000000000105bd0>"
type = 260
value = {
var = 2.1225218246347e-314
fun = 0x100103f00
}
next = 0x1001043e0
}
(dbx) quit
beta $
I definately made the change :
beta $ cat ../bison-3.8.2_sunos5.10_sparcv9.002.patch
--- examples/c/mfcalc/mfcalc.y.orig Sat Sep 25 09:10:40 2021
+++ examples/c/mfcalc/mfcalc.y Wed Oct 13 08:35:12 2021
@@ -91,7 +91,7 @@
};
/* The symbol table: a chain of 'struct symrec'. */
-symrec *sym_table;
+symrec *sym_table = NULL;
/* Put functions in table. */
static void
So this is pretty darn odd if that pointer was supposed to be initialized.
--
Dennis Clarke
RISC-V/SPARC/PPC/ARM/CISC
UNIX and Linux spoken
GreyBeard and suspenders optional
