Re: Imake on Solaris 11.2/sparc

[Jörn Clausen <joern.clausen%uni-bielefeld.de@localhost>]

Hi David!

On 04.09.16 23:43, David Holland wrote:
[..]
>   >  I will look into it when I can but that probably won't be for at least
>   >  another week.
>
> "month and a half" :(

Good enough, if you ask me ;) Thanks!

> This is now done. If you try again with imake 1.0.7nb2 on both
> platforms (this will also build a new tradcpp) you'll get a file
> .imake.cpplog in the directory where imake runs that contains an eval
> log from cpp. With luck, as long as the two machines are using the
> same version of the imake templates, diffing the two logs will produce
> useful information.

I tried this on "print/transfig". Imake is invoked four times during one 
build, i.e. there are four pairs of logs to compare.

I looked at the one generated at the top level, and it shows some 
expected differences like

 /opt/pkg-hrz/20160711/lib/X11/config/Imake.tmpl:541: Expanding macro 
ByteOrder
-/opt/pkg-hrz/20160711/lib/X11/config/Imake.tmpl:541: Substituting for 
ByteOrder: X_LITTLE_ENDIAN
-/opt/pkg-hrz/20160711/lib/X11/config/Imake.tmpl:541: Complete expansion 
for ByteOrder: X_LITTLE_ENDIAN
+/opt/pkg-hrz/20160711/lib/X11/config/Imake.tmpl:541: Substituting for 
ByteOrder: X_BIG_ENDIAN
+/opt/pkg-hrz/20160711/lib/X11/config/Imake.tmpl:541: Complete expansion 
for ByteOrder: X_BIG_ENDIAN

('-' is i86, '+' is sparc) and some unexpected

-/opt/pkg-hrz/20160711/lib/X11/config/X11.tmpl:826: #ifndef 
UseInternalMalloc: not taken
+/opt/pkg-hrz/20160711/lib/X11/config/X11.tmpl:826: #ifndef 
UseInternalMalloc: taken
+/opt/pkg-hrz/20160711/lib/X11/config/X11.tmpl:827: Defining 
UseInternalMalloc

One big chunk in the diff is, that i86 pulls in xorg.cf, which sparc 
skips completely (NB that on both platforms I use X11_TYPE=modular). 
This is part of the diff immediately in front of that block:

-/opt/pkg-hrz/20160711/lib/X11/config/svr4.cf:260: defined(XorgServer): 0
+/opt/pkg-hrz/20160711/lib/X11/config/svr4.cf:260: defined(XorgServer): 1

Again: '-' is i86 and '+' is sparc!

> Another distinct possibility is that imake is putting something wrong
> into the start file it actually passes to cpp (which is *not* the
> Imakefile) -- the new imake leaves this behind in .imake.start.c
> instead of carefully deleting it so you can't see it.

The diff here looks soothingly sane:

-#define DefaultOSName SunOS 5.11 11.2 i86pc
+#define DefaultOSName SunOS 5.11 11.2 sun4v


I'll try to get some more sense out of the diffs, but if anybody is 
interested in the logs (each file is about 16000 lines), please let me 
know. I'd be happy to share them :)

Joern

--
Jörn Clausen
Plattformen & Serverdienste
Hochschulrechenzentrum

Universität Bielefeld
Universitätsstraße 25
33615 Bielefeld
Telefon: +49 521 106-12601
E-Mail: joern.clausen%uni-bielefeld.de@localhost

http://www.uni-bielefeld.de/hrz