AL,

On 2004 Oct 11 , at 10.02, Tim Jenness wrote:

> On Sun, 10 Oct 2004, Alasdair Allan wrote:
>
>>>> I'm presuming it's a generated file as CVS knows nothing about it?
>>>
>>> ...but I'm persuming mio_err.h is supposed to be aut geneted from the
>>> mio_err.msg file by MESSGEN. Doesn't look like this is happening...
>>
>> Okay, fixed! I should of course added MIO_ERR and mio_err.h to the
>> include_MESSAGES line. Ho hum...

These were listed in noinst_MESSAGES in Makefile.am 1.3 (which was the
version after I rearranged things) -- the only difference between
include_MESSAGES and noinst_MESSAGES is that the latter aren't
installed.  $(noinst_MESSAGES) is listed in PRIVATE_INCLUDES,
correctly.

The error in that file (I discover, now I try to build it myself) is
that $(noinst_MESSAGES) isn't listed in BUILT_SOURCES.  Thus the patch
to 1.3 should be:

Basil:mag> cvs diff -r 1.3 Makefile.am
Index: Makefile.am
===================================================================
RCS file: /cvs/libraries/mag/Makefile.am,v
retrieving revision 1.3
diff -u -r1.3 Makefile.am
--- Makefile.am 14 Sep 2004 10:17:13 -0000      1.3
+++ Makefile.am 11 Oct 2004 09:41:51 -0000
@@ -54,7 +54,7 @@
  C_ROUTINES = ioc.c


-BUILT_SOURCES = $(include_MESSAGES)
+BUILT_SOURCES = $(include_MESSAGES) $(noinst_MESSAGES)


  # magtest is ADAM monolith
Basil:mag>

Do you want to make that change, Al?

> That must have got broken when Norman rearranged how .msg files are
> processed. [since it clearly used to work]

I can only guess that I must have tested this without removing enough
of the *_MESSAGES files -- sorry about that.

But now I come to rebuild everything, I discover that parsecon --
completely unrelated to this problem -- now fails to build on the Mac.
And gks currently doesn't build, for a similar-looking linking reason
(and different from that reported recently by Al, on Linux).  What
gives?  This _was_ all working the last time I tried -- I know, because
I was doing it on OSX, and would have noticed if nothing was working!

ARGHHH.  When did this stop working?  We have no clue.

Is there any progress with the nightly builds?  We don't have to wait
until everything is building: we don't have to do `make world' nightly
on all platforms, as we could just do `make /blah/manifests/xxx' for
some suitably chosen component `xxx', and inch that towards `make
world' as and when things were observed to build.  At least then we'd
know when things broke, and thus have a better chance at working out
what it was that broke them.

Norman


--
----------------------------------------------------------------------
Norman Gray  :  Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/  :  www.starlink.ac.uk