Re: [PATCH] Makefile: introduce NO_PTHREADS

[Junio C Hamano <gitster@pobox.com>]

Linus Torvalds <torvalds@linux-foundation.org> writes:

> On Sat, 15 Nov 2008, Junio C Hamano wrote:
>>
>> This introduces make variable NO_PTHREADS for platforms that lack the
>> support for pthreads library or people who do not want to use it for
>> whatever reason.  When defined, it makes the multi-threaded index
>> preloading into a no-op, and also disables threaded delta searching by
>> pack-objects.
>
> Ack. Makes sense.

Hmm, I started getting random segfaults that sometimes reproduces.  For
example, this is what I just got from "git diff --stat $some_commit".

#0  0x00002b952568b090 in strlen () from /lib/libc.so.6
#1  0x000000000044ed42 in git_checkattr (path=0x2b95284b3970 "parse-options.c",
    num=3, check=0x41000e90) at attr.c:512
#2  0x0000000000458921 in convert_to_git (
    path=0x2b95284b3970 "parse-options.c",
    src=0x2aaaaaabb000 <Address 0x2aaaaaabb000 out of bounds>, len=12594,
    dst=0x41000f30, checksafe=SAFE_CRLF_FALSE) at convert.c:578
#3  0x0000000000489ac3 in index_mem (
    sha1=0x41000ff0 "\210f�⽡x�\207�� 7R}\217\032��", buf=0x2aaaaaabb000,
    size=12594, write_object=0, type=<value optimized out>,
    path=0x2f2f2f2f2f2f2f2f <Address 0x2f2f2f2f2f2f2f2f out of bounds>)
    at sha1_file.c:2451
#4  0x0000000000489c3d in index_fd (
    sha1=0x41000ff0 "\210f�⽡x�\207�� 7R}\217\032��", fd=5,
    st=<value optimized out>, write_object=0, type=OBJ_BLOB,
    path=0x2b95284b3970 "parse-options.c") at sha1_file.c:2483
#5  0x000000000047857a in ce_modified_check_fs (ce=0x2b95284b3930,
    st=0x41001080) at read-cache.c:92
#6  0x00000000004786a2 in ie_match_stat (istate=0x71c860, ce=0x2b95284b3930,
    st=0x41001080, options=<value optimized out>) at read-cache.c:282
#7  0x0000000000497e65 in preload_thread (_data=<value optimized out>)
    at preload-index.c:46
#8  0x00002b9525964017 in start_thread () from /lib/libpthread.so.0
#9  0x00002b95256da5bd in clone () from /lib/libc.so.6
#10 0x0000000000000000 in ?? ()

I suspect that the callpath around ce_modified_check_fs() uses a buffer
obtained from path.c:get_pathname() and parallel threads stomp on each
other, but I do not have time to debug this right now (I will be on a
14-hour flight in a few hours).