bug-gnulib
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

regexp_exec thread-unsafe


From: Ludovic Courtès
Subject: regexp_exec thread-unsafe
Date: Tue, 14 May 2013 23:21:43 +0200
User-agent: Gnus/5.130005 (Ma Gnus v0.5) Emacs/24.3 (gnu/linux)

Hello,

Guile imports Gnulib ‘regexp’ module, and for some reason it gets
compiled even on glibc systems.

The problem is that Gnulib’s ‘regex_internal.h’ reads this:

  #if defined _LIBC
  # include <bits/libc-lock.h>
  #else
  # define __libc_lock_init(NAME) do { } while (0)
  # define __libc_lock_lock(NAME) do { } while (0)
  # define __libc_lock_unlock(NAME) do { } while (0)
  #endif

and ‘regexp_exec’ happily does:

    __libc_lock_lock (dfa->lock);
    if (preg->no_sub)
      err = re_search_internal (preg, string, length, start, length,
                                length, 0, NULL, eflags);
    else
      err = re_search_internal (preg, string, length, start, length,
                                length, nmatch, pmatch, eflags);
    __libc_lock_unlock (dfa->lock);

Obviously that doesn’t work well, and it caused a regression in Guile.
With MALLOC_CHECK_=2 and a Guile program, it aborts like this (sorry, I
failed to come up with a reduced test case):

--8<---------------cut here---------------start------------->8---
(gdb) bt full
#0  0x00007f793608d2b5 in raise () from /home/ludo/.guix-profile/lib/libc.so.6
No symbol table info available.
#1  0x00007f793608e738 in abort () from /home/ludo/.guix-profile/lib/libc.so.6
No symbol table info available.
#2  0x00007f79360d2afe in malloc_printerr () from 
/home/ludo/.guix-profile/lib/libc.so.6
No symbol table info available.
#3  0x00007f79360d5bf5 in realloc_check () from 
/home/ludo/.guix-profile/lib/libc.so.6
No symbol table info available.
#4  0x00007f7937a6efed in register_state (dfa=0x10872c0, address@hidden, 
address@hidden)
    at regex_internal.c:1590
        new_alloc = 8
        new_array = 0x0
        spot = 0x11981c0
        i = <optimized out>
#5  0x00007f7937a72373 in create_cd_newstate (hash=<optimized out>, context=1, 
nodes=0x12, dfa=0x1c6dce0) at regex_internal.c:1734
        i = <optimized out>
        nctx_nodes = <optimized out>
        err = <optimized out>
        newstate = 0x1c6dce0
#6  re_acquire_state_context (address@hidden, address@hidden, address@hidden, 
    address@hidden) at regex_internal.c:1554
        hash = 6
        spot = <optimized out>
        i = <optimized out>
#7  0x00007f79379bb54b in build_trtable (dfa=<optimized out>, address@hidden) 
at regexec.c:3474
        next_node = <optimized out>
        err = _REG_NOERROR
        i = 1
        j = 2
        ch = <optimized out>
        need_word_trtable = true
        elem = <optimized out>
        mask = <optimized out>
        dests_node_malloced = true
        dest_states_malloced = true
        trtable = <optimized out>
        dest_states = 0x1c688f0
        dest_states_word = 0x1c68918
        dest_states_nl = 0x1c68940
        follows = {alloc = 74, nelem = 18, elems = 0x1c5d4e0}
        dests_node = 0x1c642b0
        dests_ch = 0x1c65ab0
        acceptable = {288063250384289792, 576460743847706622, 0, 0}
        dests_alloc = 0x1c642b0
#8  0x00007f7937a77714 in transit_state (state=0x1c44820, mctx=0x7fff67bda000, 
err=0x7fff67bd9fb0) at regexec.c:2339
        trtable = <optimized out>
        ch = 116 't'
#9  check_matching (p_match_first=0x7fff67bd9fb8, fl_longest_match=true, 
mctx=0x7fff67bda000) at regexec.c:1187
        old_state = 0x1c44820
        next_char_idx = <optimized out>
        dfa = <optimized out>
        err = _REG_NOERROR
        match = 0
        match_last = <optimized out>
        next_start_idx = 0
        cur_str_idx = <optimized out>
        cur_state = 0x1c44820
        at_init_state = <optimized out>
#10 re_search_internal (preg=0x1256700, string=0x1c612d0 
"http://hydra.gnu.org/nar/0did57blsaaspj49cl3q8nnydm5pr1qg-cflow-1.4";, 
    length=67, start=0, last_start=0, stop=<optimized out>, nmatch=9, 
pmatch=0x1c63d50, eflags=0) at regexec.c:856
        err = _REG_NOERROR
        dfa = 0x10872c0
        left_lim = 0
        right_lim = 0
        incr = 1
        fl_longest_match = <optimized out>
        match_kind = 8
        match_first = 0
        match_last = <optimized out>
        extra_nmatch = 0
        sb = false
        mctx = {input = {raw_mbs = 0x1c612d0 
"http://hydra.gnu.org/nar/0did57blsaaspj49cl3q8nnydm5pr1qg-cflow-1.4";, 
            mbs = 0x1c612d0 
"http://hydra.gnu.org/nar/0did57blsaaspj49cl3q8nnydm5pr1qg-cflow-1.4";, wcs = 
0x1c63f50, 
            offsets = 0x0, cur_state = {__count = 0, __value = {__wch = 0, 
__wchb = "\000\000\000"}}, raw_mbs_idx = 0, 
            valid_len = 67, valid_raw_len = 67, bufs_len = 68, cur_idx = 2, 
raw_len = 67, len = 67, raw_stop = 67, stop = 67, 
            tip_context = 6, trans = 0x0, word_char = 0x1087378, icase = 0 
'\000', is_utf8 = 1 '\001', map_notascii = 0 '\000', 
            mbs_allocated = 0 '\000', offsets_needed = 0 '\000', newline_anchor 
= 0 '\000', word_ops_used = 0 '\000', 
            mb_cur_max = 6}, dfa = 0x10872c0, eflags = 0, match_last = 
18446744073709551615, last_node = 0, 
          state_log = 0x1c64070, state_log_top = 1, nbkref_ents = 0, 
abkref_ents = 0, bkref_ents = 0x0, max_mb_elem_len = 0, 
          nsub_tops = 0, asub_tops = 0, sub_tops = 0x0}
        fastmap = 0x0
        t = 0x0
#11 0x00007f7937a7b5c0 in rpl_regexec (preg=0x1256700, string=0x3399 <Address 
0x3399 out of bounds>, 
    address@hidden 
"http://hydra.gnu.org/nar/0did57blsaaspj49cl3q8nnydm5pr1qg-cflow-1.4";, 
nmatch=6, address@hidden, 
    pmatch=0xffffffffffffffff, address@hidden, eflags=0) at regexec.c:254
        err = _REG_NOERROR
        start = 13209
        length = 0
#12 0x00007f7937a68873 in scm_regexp_exec (rx=0x1836090, str=0x1806b00, 
start=<optimized out>, flags=0x2) at regex-posix.c:288
        status = <optimized out>
        nmatches = 9
        offset = 0
        matches = 0x1c63d50
        c_str = 0x1c612d0 
"http://hydra.gnu.org/nar/0did57blsaaspj49cl3q8nnydm5pr1qg-cflow-1.4";
        mvec = 0x4
        substr = <optimized out>
--8<---------------cut here---------------end--------------->8---

How should that be fixed?  Shouldn’t __libc_lock_unlock & co. be rebased
on top of pthread_mutex_t?

Thanks,
Ludo’.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]