Chandler 1, Mac OS X 10.6, X86-64: make DEBUG=1 world: openssl crashes, make failures

[Graham Perrin]

Re <http://chandlerproject.org/Developers/BuildingChandlerDesktop> after commanding

make DEBUG=1 world

 
whilst I was sleeping, two things occurred:

* openssl crashed five times in rapid succession between
  2009-08-29 00:10:02.208 +0100 and
  2009-08-29 00:10:03.430 +0100 (times within crash logs)

* make failed, probably 2009-08-29 00:17

Attached:

* the first of the five crash logs for
/Users/gjp22/Developer/chandler/external/openssl/openssl-0.9.8d/build_debug/apps/openssl

* Terminal output, including the environment and the time when PythonInterpreter.icns was written.

Any ideas?

Thanks
Graham

---

I should mention,

a) a much earlier topic relating to build failure before I upgraded to Mac OS X 10.6:
<http://n2.nabble.com/-tp3010831p3010831.html>; AFAIR since that time I have (amongst other things) removed and reinstalled MacPorts, so that topic may be irrelevant now

b) <https://bugzilla.osafoundation.org/show_bug.cgi?id=12860>, bug preventing me from launching a snapshot of Chandler Desktop 1.0.3.1 for Mac OS X

c) discussion with Grant, <http://chandlerproject.org/script/getIrcTranscript.cgi?messagesOnly=true&channel=chandler&date=20090828>

-- points (b) and (c) led me to attempt the build.

openssl_2009-08-29-001003_macbookpro08-centrim.crash
2009-08-29-0017.txt

[Grant Baillie]

On 28 Aug, 2009, at 22:20, Graham Perrin wrote:

>
> Re <http://chandlerproject.org/Developers/BuildingChandlerDesktop>  
> after
> commanding
> make DEBUG=1 world
> whilst I was sleeping, two things occurred:
>
> * openssl crashed five times in rapid succession between
>  2009-08-29 00:10:02.208 +0100 and
>  2009-08-29 00:10:03.430 +0100 (times within crash logs)
>
> * make failed, probably 2009-08-29 00:17
>
> Attached:
>
> * the first of the five crash logs for
> /Users/gjp22/Developer/chandler/external/openssl/openssl-0.9.8d/
> build_debug/apps/openssl

I'm wondering if some of the code that version of openssl uses to  
build itself is not 64-bit safe.

>
> * Terminal output, including the environment and the time when
> PythonInterpreter.icns was written.
>
> Any ideas?

You might try something like

make MACOSX_DEPLOYMENT_TARGET=10.5 BUILD_ROOT=$(pwd) world

I'm not sure that that will work at all: The Makefiles are building  
all kinds of unnecessary stuff, like a python interpreter, since it  
doesn't recognise Mac OS X 10.6 as a known build type.

--Grant

> I should mention,
>
> a) a much earlier topic relating to build failure _before_ I  
> upgraded to Mac
> OS X 10.6:
> <http://n2.nabble.com/-tp3010831p3010831.html>; AFAIR since that  
> time I have
> (amongst other things) removed and reinstalled MacPorts, so that  
> topic may
> be irrelevant now
>
> b) <https://bugzilla.osafoundation.org/show_bug.cgi?id=12860>, bug
> preventing me from launching a snapshot of Chandler Desktop 1.0.3.1  
> for Mac
> OS X
>
> c) discussion with Grant,
> <http://chandlerproject.org/script/getIrcTranscript.cgi?messagesOnly=true&channel=chandler&date=20090828 
> >
>
> -- points (b) and (c) led me to attempt the build.
>
> http://n2.nabble.com/file/n3540621/openssl_2009-08-29-001003_macbookpro08-centrim.crash
> openssl_2009-08-29-001003_macbookpro08-centrim.crash
> http://n2.nabble.com/file/n3540621/2009-08-29-0017.txt 
> 2009-08-29-0017.txt
> --
> View this message in context: http://n2.nabble.com/Chandler-1-Mac-OS-X-10-6-X86-64-make-DEBUG-1-world-openssl-crashes-make-failures-tp3540621p3540621.html
> Sent from the chandler-dev mailing list archive at Nabble.com.
> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
>
> Open Source Applications Foundation "chandler-dev" mailing list
> http://lists.osafoundation.org/mailman/listinfo/chandler-dev

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev

[Graham Perrin]

Grant Baillie wrote:

I'm wondering if some of the code that version of openssl uses to build itself is not 64-bit safe.

I wondered the same thing. Saw the
0.9.8d at <http://svn.osafoundation.org/chandler/trunk/external/openssl/Makefile> and
0.9.8k (latest) at <http://www.openssl.org/source/>.

Today, experimenting with two versions 0.9.7m and 0.9.8k via MacPorts,† I found that the earlier version was fine but that the latest version 0.9.8k required a clean before it would build. Notes at <http://pastebin.ca/1550494>. Does that offer any clue?

Do we need a version later than 0.9.8d for Chandler 1.x to build?

I wonder...

† <http://trac.macports.org/browser/trunk/dports/devel/openssl97/Portfile>
  <http://trac.macports.org/browser/trunk/dports/devel/openssl/Portfile>

[Grant Baillie]

On 1 Sep, 2009, at 10:13, Graham Perrin wrote:

>
>
> Grant Baillie wrote:
>>
>> I'm wondering if some of the code that version of openssl uses to  
>> build
>> itself is not 64-bit safe.
>>
>
> I wondered the same thing. Saw the
> 0.9.8d at
> <http://svn.osafoundation.org/chandler/trunk/external/openssl/ 
> Makefile> and
> 0.9.8k (latest) at <http://www.openssl.org/source/>.
>
> Today, experimenting with two versions 0.9.7m and 0.9.8k via  
> MacPorts,† I
> found that the earlier version was fine but that the latest version  
> 0.9.8k
> required a clean before it would build. Notes at
> <http://pastebin.ca/1550494>. Does that offer any clue?

I think the error there:

gjp22% sudo port install openssl
--->  Computing dependencies for openssl
--->  Building openssl
Error: Target org.macports.build returned: shell command " cd "/opt/
local/var/macports/build/
_opt_local_var_macports_sources_rsync
.macports.org_release_ports_devel_openssl/work/openssl-0.9.8k" && /usr/
bin/make all " returned error 2
#
Command output: ld: warning: in /opt/local/lib/libz.dylib, file is not  
of required architecture

indicates to me that you have an old (i.e. non-64 bit) macports  
installation. You probably want to move aside /opt and start again.

>
> Do we need a version later than 0.9.8d for Chandler 1.x to build?

I'm not sure about this.

--Grant

>
> I wonder...
>
> † <http://trac.macports.org/browser/trunk/dports/devel/openssl97/Portfile 
> >
>   <http://trac.macports.org/browser/trunk/dports/devel/openssl/Portfile 
> >
> --
> View this message in context: http://n2.nabble.com/Chandler-1-Mac-OS-X-10-6-X86-64-make-DEBUG-1-world-openssl-crashes-make-failures-tp3540621p3561758.html
> Sent from the chandler-dev mailing list archive at Nabble.com.
> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
>
> Open Source Applications Foundation "chandler-dev" mailing list
> http://lists.osafoundation.org/mailman/listinfo/chandler-dev

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev

[Graham Perrin]

Grant Baillie wrote:

… indicates to me that you have an old (i.e. non-64 bit) macports installation. You probably want to move aside /opt and start again.

I removed my old
/opt/local
(nothing I wanted from there), reinstalled MacPorts etc., got as far as  
make DEBUG=1 world
and encountered the same errors:

cc1: error: unrecognized command line option "-Wno-long-double"
make[2]: *** [Modules/python.o] Error 1
make[1]: *** [build] Error 2
make: *** [python] Error 2

Selected details pasted to <http://pastebin.ca/1552370> for a year. In the opening lines there I'm reminded that I had to create some symlinks in
/opt/local/bin
and now looking at my PATH:

PATH=/opt/local/bin:/opt/local/sbin:/opt/local/bin:/opt/local/sbin:/opt/local/bin:/opt/local/sbin:/Users/gjp22/Developer/depot_tools:/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin:/usr/X11/bin

— I wonder whether it's good to have the paths for MacPorts coming before the more standard paths.

(Using Mac OS X 10.5.x, I struggled with environment settings to get the PATH as I wanted it, AFAIR it never worked out as I wanted. Off-topic from Chandler, I suspect that a default upgrade from 10.5.8 to 10.6 has worked some magic.)