Hi Simon.

| > mthomas     2004/03/09 23:02:47 PST
| >
| >   Modified files:        (Branch: ghc-6-2-branch)
| >     .                    Makefile
| >   Log:
| >   Further tweak doc build for machines without the tools.
| >
| >   Revision  Changes    Path
| >   1.100.2.2 +4 -0      fptools/Makefile
|
| I don't understand why you need the extra check in the install-docs rule
| - what goes wrong without this?  I can see why you might need it in the
| binary-dist rule, although binary dists don't work on Windows, do they?

On Windows the bindist target does produce a binary distribution directory
but the actual distribution doesn't work on Windows; at the very least
because it depends on a shell script.

When I first modified the nightly build system to build various utilities I
followed and extended the example of the MinGW ghc "rejig" script
"distrib/prep-bin-dist-mingw" on the assumption that it was the "done thing
over there" and I didn't want to produce something which would disrupt
Sigbjorn's packaging effort.

That script and those I made for each particular utility (all living in the
"fptools/distrib" sub directory) require a Unix style bindist directory to
start from, copying the required elements into a single directory tree
comprising the compiler, utilities and libraries which I then zip up.

The problem is that the Makefile overrides the setting of SGMLDocWays as an
argument to make adding the values html and ps through

    "BINDIST_DOC_WAYS = html ps":

Consequently:

------------------------------------------------------------------------
$ make -C haddock install-docs SGMLDocWays="html"
make: Entering directory `/c/cvs/stable/i386-unknown-mingw32/haddock'
------------------------------------------------------------------------
===fptools== Recursively making `install-docs' in src html doc ...
PWD = /c/cvs/stable/i386-unknown-mingw32/haddock
------------------------------------------------------------------------
------------------------------------------------------------------------
==fptools== make install-docs - --unix -wr;
 in /c/cvs/stable/i386-unknown-mingw32/haddock/src
------------------------------------------------------------------------
make[1]: Nothing to be done for `install-docs'.
------------------------------------------------------------------------
==fptools== make install-docs - --unix -wr;
 in /c/cvs/stable/i386-unknown-mingw32/haddock/html
------------------------------------------------------------------------
make[1]: Nothing to be done for `install-docs'.
------------------------------------------------------------------------
==fptools== make install-docs - --unix -wr;
 in /c/cvs/stable/i386-unknown-mingw32/haddock/doc
------------------------------------------------------------------------
../../glafp-utils/docbook/db2html -d
c:/cvs/stable/i386-unknown-mingw32/docs/fpt
ools-both.dsl   haddock.sgml
CATALOG file not set up; see installation guide for details.
make[1]: *** [haddock.html] Error 1
make: *** [install-docs] Error 1
make: Leaving directory `/c/cvs/stable/i386-unknown-mingw32/haddock'
------------------------------------------------------------------------

I imagine this schemozzle can be sorted out by rearranging the make install
target for Windows as has been proposed and discussed previously - I just
haven't attacked that problem yet as I first wanted to get the basic nightly
infrastructure under control so that coordinated compiler and utility
releases can be made automatically.

Anyway, all input on this is welcome and I appreciate the interest you have
shown in what I am doing - I should perhaps have discussed my plans more
openly early in the piece.

Cheers

Mike Thomas.


_______________________________________________
Cvs-fptools mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/cvs-fptools

Reply via email to