patch, 5 partsHow I compile on IRIX 6.5 with the MIPSpro compiler and ksh

From: Brandon Casey <drafnel@gmail.com>

This email describes how I have successfully compiled git on IRIX6.5 with the native MIPSpro compiler version 7.4.4m and successfully run nearly all of the tests using the Korn shell.

A config.mak file is provided at the end of this email.

The GIT_SKIP_TESTS environment variable was used to skip tests which still fail. The number of tests which must be skipped has been greatly reduced. When I sent this email last year, 23 individual tests were skipped plus all of t3901. With current master, only 13 tests must be skipped. The remaining tests which fail are due to the installed iconv lacking support for the tested languages.

Two lingering concerns:
  1) Building with -Ofast produces an executable which segfaults and can not
     pass the tests.
  2) Building with mmap support produces an executable which segfaults on
     t1400-update-ref when it calls

         'git rev-parse --verify "master@{May 25 2005}"'

     Previously, this fault was avoided when THREADED_DELTA_SEARCH was set,
     but there is a warning during compilation when THREADED_DELTA_SEARCH is
     enabled that suggests Variable Length Arrays should not be used when
     threads are used.  So it is left unset, and now NO_MMAP is set.

     From my notes:
     I tracked the mmap issue down to the mmap in refs.c:read_ref_at().  The
     mmap succeeds, and the loop is executed properly until rec is rewound
     into the first line (reflog entry) of the file. I don't know why it
     fails, the code looks correct. I can't reproduce it with a similarly
     structured stand-alone test program.  For the mentioned test that fails
     above, if the rev-parse argument is changed to something that does not
     reference the first entry in the log file, then it will succeed
     e.g. "master@{3}".

I have no debugger.

Otherwise, I've been using this executable somewhat and have not experienced any problems. Maybe it will be useful to others.

Are there actually any IRIX users out there?

These patches are all submitted For-Your-Information only. Junio, feel free to apply any of them or none, and to add my Signed-off-by.

-brandon

Brandon Casey (5):
  unpack-trees.c: work around run-time array initialization flaw on
    IRIX 6.5
  git-compat-util.h: adjust for SGI IRIX 6.5
  Makefile: add NEEDS_LIBGEN to optionally add -lgen to compile
    arguments
  Makefile: add section for SGI IRIX 6.5
  test-lib.sh: work around ksh's trap shortcomings

 Makefile          |   18 ++++++++++++++++++
 configure.ac      |    6 ++++++
 git-compat-util.h |    3 ++-
 t/lib-git-svn.sh  |   15 ++++++++++-----
 t/test-lib.sh     |    3 ++-
 unpack-trees.c    |    2 +-
 6 files changed, 39 insertions(+), 8 deletions(-)

--->8--- config.mak --->8---

GIT_SKIP_TESTS := \
   t3900.1[129] t3900.2[0234] \
   t5100.5 t5100.1[06] \
   t8005.[234]
export GIT_SKIP_TESTS

SHELL_PATH = /bin/ksh TAR = gtar CC = c99 CFLAGS = -n32 -O2 NO_C99_FORMAT = 1 NO_CURL = 1 NO_OPENSSL = 1 NO_TCLTK = 1

The c99 MIPSpro Compiler version 7.4.4m on IRIX 6.5 does not properly
initialize run-time initialized arrays.  An array which is initialized with
fewer elements than the length of the array should have the unitialized
elements initialized to zero.  This compiler only initializes the remaining
elements when the last element is a static parameter.  So work around it
by adding a "NULL" initialization parameter.
---
 unpack-trees.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

Don't define _XOPEN_SOURCE Do define _SGI_SOURCE

Defining _XOPEN_SOURCE prevents many of the common functions and macros
from being defined. _Not_ setting _XOPEN_SOURCE, and instead setting
_SGI_SOURCE, provides all of the XPG4, XPG5, BSD, POSIX functions and
declarations, _BUT_ provides a horribly broken snprintf(). SGI does have
a working snprintf(), but it is only provided when _NO_XOPEN5 evaluates
to zero, and this only happens if _XOPEN_SOURCE is defined which, as
mentioned above, prevents many other common functions and defines.
---
 git-compat-util.h |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

From: Brandon Casey <drafnel@gmail.com>

Commit 003b33a8 recently added a call to basename(). On IRIX 6.5, this function resides in libgen and -lgen is required for the linker.

Update configure.ac too. ---

configure.ac changes are untested.

-brandon

 Makefile     |    5 +++++
 configure.ac |    6 ++++++
 2 files changed, 11 insertions(+), 0 deletions(-)

---
 Makefile |   13 +++++++++++++
 1 files changed, 13 insertions(+), 0 deletions(-)

From: Brandon Casey <casey@argon.nrlssc.navy.mil>

In ksh, if trap is called within a function with 0 or EXIT as its signal, then the trap will be executed at the time the function returns. This causes a problem in the test functions since 'trap - EXIT' is called within the test_done function in order to remove the trap which calls die() on exit. This means trap has to be called from the scripts top-level. Do so by renaming the test_done function to test_done_func, and then, remove the trap and call test_done_func from within a new alias named test_done.

Additionally, there is some strangeness with respect to aliases and sourced
script files; the alias is not available within the sourced file. So call
'trap - EXIT' directly in lib-git-svn.sh before calling the test_done_func
function.
---
 t/lib-git-svn.sh |   15 ++++++++++-----
 t/test-lib.sh    |    3 ++-
 2 files changed, 12 insertions(+), 6 deletions(-)

From: Brandon Casey <drafnel@gmail.com>

---

Here's the update for the 64-bit version of IRIX. This has been tested to the same extent that the 32-bit version has been tested. The 64-bit test system is a little bit older than the 32-bit system, so a few more tests are skipped (related to a too old gtar), and NO_STRLCPY and NO_DEFLATE_BOUND macros are set.

-brandon

 Makefile |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)

Brandon Casey <casey@nrlssc.navy.mil> writes:

I thought at some point Dscho was suffering from IRIX but perhaps he got out of it?

Except for 5/5 all of them looked trivially safe and none of them looked iffy. Will queue, possibly with minor tweaks.