nmh-workers
[Top] [All Lists]

Re: [Nmh-workers] GCC 8 pre-releases have escaped...

2018-02-13 03:56:16
Hi Ken,

Alright, I pushed a change to master and 1.7-release that eliminates
that call to free(). Ralph, does that solve your problem?

With 1.7-branchpoint-495-g19c69cf3, valgrind 1:3.13.0-1ubuntu3 on one
machine moans.

    Source and destination overlap in memcpy_chk(0x1ffefeedc0, 0x1ffefeedc0, 22)
       at 0x4C37660: __memcpy_chk (in 
/usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
       by 0x11BAA4: memmove (string_fortified.h:40)
       by 0x11BAA4: fmt_scan (fmt_scan.c:756)
       by 0x115B78: build_form (forwsbr.c:193)
       by 0x10F2BF: main (forw.c:455)
    FAIL: test/forw/test-forw-format

    Source and destination overlap in memcpy_chk(0x1ffeff3060, 0x1ffeff3061, 18)
       at 0x4C37660: __memcpy_chk (in 
/usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
       by 0x11CA74: memmove (string_fortified.h:40)
       by 0x11CA74: fmt_scan (fmt_scan.c:756)
       by 0x1100C9: replout (replsbr.c:225)
       by 0x10E99E: main (repl.c:470)
    first named test failure: -cc me

But they make no sense since memmove(3)'s point is the regions may
overlap.

The same 1.7-branchpoint-495-g19c69cf3 on another machine with valgrind
3.13.0-6 running just those two tests under valgrind is silent.

Maximum resident set size.  (I forgot to say previously that I've no
swap for any of these measurements.)

    $ for n in 1 10 100 1000 10000; do
    >     \time -f %M uip/scan -width 80 last:$n >/dev/null
    > done
    3872
    3808
    4012
    4056
    4180
    $

Those previous troublesome emails.

    $ valgrind uip/scan -forma '%<{date}d%>%<(mymbox{from})f%<(mymbox{to})t%>%>'
    dft
    dft
    dft
    dft
    dft
    d
    ==978== HEAP SUMMARY:
    ==978==     in use at exit: 58,467 bytes in 206 blocks
    ==978==   total heap usage: 2,845 allocs, 2,639 frees, 432,238 bytes 
allocated
    ==978== 
    ==978== LEAK SUMMARY:
    ==978==    definitely lost: 142 bytes in 3 blocks
    ==978==    indirectly lost: 40 bytes in 3 blocks
    ==978==      possibly lost: 56 bytes in 1 blocks
    ==978==    still reachable: 58,229 bytes in 199 blocks
    ==978==         suppressed: 0 bytes in 0 blocks

So all looks good to me.

-- 
Cheers, Ralph.
https://plus.google.com/+RalphCorderoy

-- 
Nmh-workers
https://lists.nongnu.org/mailman/listinfo/nmh-workers

<Prev in Thread] Current Thread [Next in Thread>