[lmi] Building lmi under Linux

[Vadim Zeitlin]

 Hello,

 As you know, we currently build lmi under Linux using autotools-based
build system. This works pretty well for lmi itself, but it doesn't take
care of all the prerequisites, such as Boost, xmlwrapp and wxWidgets, which
I had been building manually, using their own build systems until now.

 In order to make it simpler for the others to build lmi out of the box,
we've now made a trivial shell script which simply performs all these steps
(basically, "mkdir build-gtk && cd $_ && ../configure && make install") for
all the dependencies and lmi itself in a row. This works, but is not ideal
because it overlaps with install_msw.sh, but doesn't do everything that
the other script does.

 Hence my first question: should we try to fully duplicate install_msw.sh?
I.e. do we want to have install_linux.sh which exactly mirrors the existing
script or maybe even a single install_lmi.sh which works under any system
and respects LMI_TRIPLET specified in the environment? There are clearly
some differences between the platforms, but there is also quite some
overlap.

 One of the consequences of this decision is that if we want to reuse
install_msw.sh logic under Linux, we need to get rid of autotools-based
system and use lmi makefiles there too. Personally, I would greatly miss
configure+make, of course, but I think there is no chance that you'd accept
to use under MSW and having the same build scripts but ending up using 2
different build systems at the end clearly doesn't make much sense.

 To summarize, I think there are 2 reasonable ways of building lmi under
Linux:

1. The usual one, with installing all the dependencies manually and then
   running lmi configure, which will pick them up and build lmi itself.
   For simplicity, this could be done by a single _simple_ wrapper script.

2. The lmi one, with all the extra steps done by install_msw.sh and the
   other install_xxx files used by it, which is far from simple, but has
   the advantage of being comprehensive.

 Which one should we pursue? I guess your preference would be (2), but I'd
prefer to be sure before embarking on all the changes that will be needed
to replace the existing script doing (1).

 Thanks in advance,
VZ

pgpP_TvGRUrNV.pgp
Description: PGP signature