Re: Problems with git 1.8.5.3 on HP-UX 11.11

2014-03-28 Thread Jeff King
On Fri, Mar 28, 2014 at 03:43:29PM -0400, Eric Sunshine wrote:

> On Fri, Mar 28, 2014 at 3:01 PM, Jeff King  wrote:
> > On Fri, Mar 28, 2014 at 11:09:14AM -, Gerhard Grimm wrote:
> >> git submodule init
> >>
> >> fails with the output
> >>
> >> Assertion failed: err == REG_ESPACE, file compat/regex/regexec.c, line 
> >> 1096
> >> No submodule mapping found in .gitmodules for path 'module'
> >
> > The regexes we use here are not particularly complicated. So either
> > there is a bug (but nobody else has reported anything on any other
> > platform) or your system regex library has some problem with what we are
> > feeding it. The simplest solution may be to compile with:
> >
> >   NO_REGEX=YesPlease
> >
> > which will build and use the glibc implementation in compat/regex
> > instead.
> 
> Based upon the assertion-failure message, it looks like he's already
> using compat/regex.

Heh, I didn't even notice that. I just looked at all of the libc calls
at the top of the backtrace, but of course that is just from assert() on
up.

So now it seems doubly odd to me, since it is running the same regex
library that is used elsewhere.

-Peff
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html


Re: Problems with git 1.8.5.3 on HP-UX 11.11

2014-03-28 Thread Eric Sunshine
On Fri, Mar 28, 2014 at 3:01 PM, Jeff King  wrote:
> On Fri, Mar 28, 2014 at 11:09:14AM -, Gerhard Grimm wrote:
>> git submodule init
>>
>> fails with the output
>>
>> Assertion failed: err == REG_ESPACE, file compat/regex/regexec.c, line 
>> 1096
>> No submodule mapping found in .gitmodules for path 'module'
>
> The regexes we use here are not particularly complicated. So either
> there is a bug (but nobody else has reported anything on any other
> platform) or your system regex library has some problem with what we are
> feeding it. The simplest solution may be to compile with:
>
>   NO_REGEX=YesPlease
>
> which will build and use the glibc implementation in compat/regex
> instead.

Based upon the assertion-failure message, it looks like he's already
using compat/regex.
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html


Re: Problems with git 1.8.5.3 on HP-UX 11.11

2014-03-28 Thread Jeff King
On Fri, Mar 28, 2014 at 11:09:14AM -, Gerhard Grimm wrote:

> So I examined the git source package and found that the author of the
> HP-UX port forgot to set
> 
> PTHREAD_CFLAGS=-mt
> 
> in config.mak.autogen to enable threading.

You probably want to place such manual settings in config.mak. If you
use the "./configure" script, it will overwrite config.mak.autogen.

> git submodule init
> 
> fails with the output
> 
>     Assertion failed: err == REG_ESPACE, file compat/regex/regexec.c, line 
> 1096
>     No submodule mapping found in .gitmodules for path 'module'
> 
> and the stacktrace of the resulting core dump is
> 
> #0  0xc020ced0 in kill+0x10 () from /usr/lib/libc.2
> #1  0xc01a7f84 in raise+0x24 () from /usr/lib/libc.2
> #2  0xc01e9308 in abort_C+0x160 () from /usr/lib/libc.2
> #3  0xc01e9364 in abort+0x1c () from /usr/lib/libc.2
> #4  0xc0176998 in _assert+0x178 () from /usr/lib/libc.2
> #5  0x205fa0 in check_matching+0x290 ()
> #6  0x2053b8 in re_search_internal+0x128 ()
> #7  0x204ac0 in regexec+0xc8 ()
> #8  0x4da40 in collect_config+0x60 ()
> #9  0x108b30 in get_value+0xd8 ()
> [...]

The regexes we use here are not particularly complicated. So either
there is a bug (but nobody else has reported anything on any other
platform) or your system regex library has some problem with what we are
feeding it. The simplest solution may be to compile with:

  NO_REGEX=YesPlease

which will build and use the glibc implementation in compat/regex
instead.

-Peff
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html


Re: Problems with git 1.8.5.3 on HP-UX 11.11

2014-03-28 Thread Thomas Ackermann
Gerhard Grimm  detec.com> writes:

> 
> In order to set up automated builds and tests of the CMake toolchain 
(www.cmake.org) on HP-UX 11.11 (hppa)
> and 11.23 (ia64), I needed to install git on those platforms.
> The latest binary package available from hpux.connect.org.uk is version 
1.8.5.3, which I installed with
> all of its dependencies.

Did you try to build the most current version v1.9.1 by using
autoconf as described in 'INSTALL'?

---
Thomas




--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html


Problems with git 1.8.5.3 on HP-UX 11.11

2014-03-28 Thread Gerhard Grimm
In order to set up automated builds and tests of the CMake toolchain 
(www.cmake.org) on HP-UX 11.11 (hppa) and 11.23 (ia64), I needed to install git 
on those platforms.
The latest binary package available from hpux.connect.org.uk is version 
1.8.5.3, which I installed with all of its dependencies.
When trying to set up the CMake build, I ran into the first problem:

$ git pull origin
error: cannot create thread: Function is not available
fatal: fetch-pack: unable to fork off sideband demultiplexer

So I examined the git source package and found that the author of the HP-UX 
port forgot to set

PTHREAD_CFLAGS=-mt

in config.mak.autogen to enable threading. I added this setting and rebuilt 
git. On 11.23, everything was fine now - no further issues.
On 11.11 though, git now crashed with a Bus Error. Some debugging showed that 
this was due to a multithreading issue - obviously some dependency library has 
not been built as reentrant code. To fix this, I disabled threading by setting

PTHREAD_CFLAGS=
NO_PTHREADS=YesPlease

in config.mak.autogen and rebuilt git again. After that, "git pull" and "git 
fetch" worked correctly and I could proceed to set up the CMake build and test.
Alas, the CMake tests include a test case "CTest.UpdateGIT" that creates a git 
repository, creates a submodule, imports some content and attempts to check out 
a revision. At that point, the command

git submodule init

fails with the output

    Assertion failed: err == REG_ESPACE, file compat/regex/regexec.c, line 1096
    No submodule mapping found in .gitmodules for path 'module'

and the stacktrace of the resulting core dump is

#0  0xc020ced0 in kill+0x10 () from /usr/lib/libc.2
#1  0xc01a7f84 in raise+0x24 () from /usr/lib/libc.2
#2  0xc01e9308 in abort_C+0x160 () from /usr/lib/libc.2
#3  0xc01e9364 in abort+0x1c () from /usr/lib/libc.2
#4  0xc0176998 in _assert+0x178 () from /usr/lib/libc.2
#5  0x205fa0 in check_matching+0x290 ()
#6  0x2053b8 in re_search_internal+0x128 ()
#7  0x204ac0 in regexec+0xc8 ()
#8  0x4da40 in collect_config+0x60 ()
#9  0x108b30 in get_value+0xd8 ()
#10 0x108efc in git_parse_source+0x1bc ()
#11 0x10ac70 in do_config_from+0x70 ()
#12 0x10ad3c in git_config_from_file+0x8c ()
#13 0x10b274 in git_config_with_options+0x84 ()
#14 0x4dd6c in get_value+0x224 ()
#15 0x4eed4 in cmd_config+0x744 ()
#16 0x17150 in run_builtin+0x110 ()
#17 0x1739c in handle_internal_command+0xcc ()
#18 0x174fc in run_argv+0x2c ()
#19 0x17724 in main+0x194 ()

Since I'm no git expert (I'm not even a regular git user in fact), there's 
nothing left for me to do except asking for help...
Please CC me (gerhard dot grimm at detec dot com) with any replies since I'm 
not subscribed to the list. Thank you!

Best regards,

Gerhard
This e-mail message has been scanned and cleared by Postini / Google Message 
Security and the UNICOM Global security systems. This message is for the named 
person's use only. If you receive this message in error, please delete it and 
notify the sender. 

--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html