git 1.6.1 on AIX 5.3

I tried building git 1.6.1 on AIX 5.3 as an "out of tree" build and it does not seem to be set up to do out of tree builds. If that is not true, please let me know.

The install process wants to call install with a -d option. AIX has two install programs but they are pretty old -- neither takes a -d option.

Is there a GNU install program I can get? I've not been able to locate one.

The curious thing is I installed git 1.5.6.4 on a similar AIX system a few months ago and succeeded.

Last -- just so I know for future reference, is this list a 'text only email' list?

Thank you for your help, Perry Ease Software, Inc. ( http://www.easesoftware.com )

Low cost SATA Disk Systems for IBMs p5, pSeries, and RS/6000 AIX systems

[cc-ing Mike Ralphson, our local AIX expert]

On Mon, Jan 26, 2009 at 02:02:15PM -0600, Perry Smith wrote:

It's in GNU coreutils:

  http://www.gnu.org/software/coreutils/

I don't know what Mike uses to install on AIX; you can see his config setup here:

  http://repo.or.cz/w/git/gitbuild.git?a=tree;f=mr/aix;hb=platform

but I don't see any override of install.

If you mean "no html", then yes, it is absolutely text only.

-Peff

2009/1/26 Jeff King <peff@peff.net>:

Bless you!

I've got the AIX Toolbox for Linux applications[1] installed and ahead of /bin and /usr/bin on my PATH. Beware that some of these don't function as well as the stock AIX utilities and should normally be removed. I've posted about this on what is ostensibly a blog but which is really just a post-it note I'm less likely to lose[2].

Failing that, many programs which are primarily configured using autoconf will ship with an install shell script which you can co-opt, some apache stuff does too[3].

Any other questions, don't hesitate to ask.

Mike

[1] http://www-03.ibm.com/systems/power/software/aix/linux/index.html

[2] http://mermade.blogspot.com/2008/04/aix-toolbox.html

[3] http://svn.apache.org/repos/asf/tcl/websh/trunk/src/unix/install-sh

On Jan 26, 2009, at 4:32 PM, Mike Ralphson wrote:

Thanks guys. I picked up coreutils version 7. I didn't install them but just moved ginstall over to /usr/local/bin.

A few other comments: I had to add in the --without-tcltk flag. I don't have tcl installed but the config did not autodetect that it was not present.

I can't tell if make test is happy or not. The output looks like its happy but the exit code is 2.

Below is my "configure" script if anyone is interested.

#!/usr/bin/env bash

export CONFIG_SHELL=/usr/local/bin/bash export LDFLAGS='-L/usr/local/lib -L/usr/local/ssl/lib' export CFLAGS='-I/usr/local/include -I/usr/local/ssl/include' export CC=gcc echo CONFIG_SHELL set to ${CONFIG_SHELL}

${CONFIG_SHELL} ../../src/git-1.6.1/configure --without-tcltk

# # Note that to install you need to do: # make INSTALL=ginstall install # to use GNU's install program

On Mon, Jan 26, 2009 at 04:57:16PM -0600, Perry Smith wrote:

Hmm. It looks like we respect --with[out]-tcltk, and without it we always say "just use wish from the PATH" without detecting whether it actually exists:

    # No Tcl/Tk switches given. Do not check for Tcl/Tk, use bare
    # 'wish'.
    TCLTK_PATH=wish
    AC_SUBST(TCLTK_PATH)

I'm sure the fix would be something along the lines of

    if which wish; then
      TCLTK_PATH=wish
    else
      NO_TCLTK=yes
    fi

but I know for fact that is not portable and that there must be some special autoconf way of doing the same thing.

That doesn't sound very happy. You should see either a "command failed" error from make, or some results like:

   '/bin/sh' ./aggregate-results.sh test-results/t*-*
   fixed   1
   success 4026
   failed  0
   broken  3
   total   4030

where "broken" is OK (it is a test that is marked as "we know this is broken currently, but ideally it would be fixed in the future") but "failed" is a problem.

But I believe unless you are using "make -k", that it won't even aggregate the results if something fails, and you should just see make complaining about the failed test script.

-Peff

2009/1/26 Perry Smith <pedzsan@gmail.com>:

Yup, I usually build with NO_TCLTK=YesPlease in my config.mak, which you can see from the link Peff posted.

When you said 'out of tree builds' I thought you meant using the AIX defaults in the Makefile without running ./configure, but it seems not, you meant is a clone of the tree buildable by itself or are only release snapshots buildable. Sorry.

./configure is (deliberately) a second-class citizen in the world of git, and may still get you a slightly suboptimal build compared with the defaults on platforms such as AIX... e.g. it doesn't test for performance-related switches such as INTERNAL_QSORT. If you have run configure, there's some file you need to blow away to get back to a non-autoconf world... is it config.mak.autogen?

Despite that, your problem with --without-tcltk falling back to wish, but not falling back if that isn't installed does look like something we should fix, as per Peff's mail.

I build with SHELL_PATH={path}/bash as well. If I don't, the test suite exits after t0000-basic.sh with an unexpected error despite passing all the tests.

gmake -C t/ all
gmake[1]: Entering directory `/usr/local/src/gitbuild/t'
rm -f -r test-results
gmake aggregate-results-and-cleanup
gmake[2]: Entering directory `/usr/local/src/gitbuild/t'
*** t0000-basic.sh ***
* passed all remaining 40 test(s)
FATAL: Unexpected exit with code 0
gmake[2]: *** [t0000-basic.sh] Error 1
gmake[2]: Leaving directory `/usr/local/src/gitbuild/t'
gmake[1]: *** [all] Error 2
gmake[1]: Leaving directory `/usr/local/src/gitbuild/t'
gmake: *** [test] Error 2

Is that what you're seeing? There's many more test scripts than that 8-)

With GIT_SKIP_TESTS='t3900 t3901 t5100.[12] t8005' (to omit some cases not handled by the version of iconv I have access to on all the AIX boxes I deploy to) the test suit runs to completion on AIX 5.3 for me.

BTW Are you running AIX's make or GNU make?

Mike

On Jan 27, 2009, at 4:10 AM, Mike Ralphson wrote:

Sorry... I thought "out of tree" was a universal term.

Just to be sure we are on the same page. My directory structure has a top/src/git-1.6.1 and top/build/git.1.6.1. The src/git-1.6.1 is the tar ball. The build/git-1.6.1 starts out empty. I cd into it and then do: ../../src/git-1.6.1/configure <options> After this completes, you can do "make".

About 90% of the open source configure / autoconf code out there can do this with. The other 10% you can not. I like it because when things die, its easier to grep around the source tree and I blow away the build directory and start back over and I know that I'm starting fresh.

Yea, part of my larger quest is to start a web site to provide AIX builds of open source images as installp images. I have not added iconv to the mix yet because I fear bad interactions with GNU's and AIX's. In theory, AIX's commands should use AIX's library but it would not surprise me if there are a few AIX applications that are not built properly.

yes -- GNU's make and gcc 4.3.1

I just tried setting SHELL_PATH (and exporting it) and the make test gets into the second batch and then fails two tests. I assume the GIT_SKIP_TESTS is done:

make GIT_SKIP_TESTS=....test

I get further. But now test 10 of t0001.sh fails because test_cmp can not be found.

Is that a GNU tool?  (I didn't see it in git or coreutils.)

(I'll go searching for it...)

Thank you, Perry

On Tue, Jan 27, 2009 at 07:35:00PM -0600, Perry Smith wrote:

I don't see how this would work without automake support, which git does not have. The configure script just generates a config.mak.autogen file, which is included by the Makefile. So you have no Makefile in your build directory.

Another way of solving the same problem:

  cd untarred-sources
  git init
  git add .
  git commit -m 'pristine sources'

Now you can use "git clean" to clean up cruft, not to mention the usual git stuff like tracking any changes you've made and submitting any patches upstream.

It's a shell function defined in test-lib.sh (which is sourced by all of the test scripts).

-Peff

Jeff King <peff@peff.net> writes:

... nor want to have ;-).

On Tue, Jan 27, 2009 at 11:37:21PM -0800, Junio C Hamano wrote:

Heh. Yes, in case there was any confusion: I don't want my statement in any way to be construed as a suggestion to support automake.

I would not be opposed to it if it somehow enhanced some users' experience without bothering people who didn't want to touch it (like the way that autoconf support is implemented). But I don't see how that would be possible.

-Peff

On Jan 28, 2009, at 1:42 AM, Jeff King wrote:

Thats fine. I didn't know if this was a "known" situation or not so I just mentioned it.

As I mentioned, most open source code supports this style of build. I have no idea what it would take to get it to work.

2009/1/28 Perry Smith <pedzsan@gmail.com>:

No, my mistake, I didn't think about what you meant.

Even if you're only building git, unless disc space is tight, I'd probably recommend cloning the git.git repository, then you can either use make clean or git clean, plus it makes updating to latest/arbitrary versions easier. Anyway, I don't think your setup is wrong.

A pity package distribution for AIX is so fragmented, but when IBM had the ball, they dropped it. Mail me if you want links to other existing sites who might appreciate contributed builds.

It's documented in test/README

As Peff said, there's a wrapper around either diff -u or a user supplied comparison tool (e.g. cmp) set in GIT_TEST_CMP. Setting it to cmp works fine on AIX if you don't have a diff which takes -u on your PATH.

As (at the moment) you need bash, gcc, gmake and ginstall to build, test and (maybe) run all parts of git on AIX, I haven't set up a clean environment to try and flush out any other quirks to get git to build/test/run on a stock AIX with no GNU tools installed. For obvious reasons such AIX machines being used for dev purposes are (I believe) rare.

Mike