On 16/02/2009 13:28, bjoern michaelsen - Sun Microsystems - Hamburg
Germany wrote:
Jussi Pakkanen wrote:
Compared to dmake, CMake has the following advantages for OOo.
- actively being developed
- used widely and thus known by lots of people
- native support for all major platforms and IDEs
- cross-platform autoconf replacement
- straightforward syntax, no shell magicks required (but you can use
them if you want to)
Other nice features can be found here:
http://www.cmake.org/Wiki/Really_Cool_CMake_Features
Actually, cmake has way to much features ;-)
The key requirements for a build system for OOo is that it has:
- few dependencies
- portable
yes.
- small
- limited to the set of features absolutely needed
(every additional feature is a lockin)
i'm not sure why features would be a problem:
- whether the features are used is more a question of requirements
- with a more primitive system, many of the required features must
be built (and maintained) by hand
- we shouldn't switch our build system every other year; if we switch
from dmake, then whatever we switch to should be good enough for
the next 15 years
While dmake might be obscure, it fits these requirements pretty well,
and more is lost than gained IMHO by moving to CMake.
here's 2 requirements that dmake does not meet:
- familiar to new developers
- maintenance is Someone Else's Problem, not a drag on OOo resources
CMake might have something going for it as a replacement for the mess
that is autotools, however thats not an issue with OOo.
well, i guess noone would disagree with your assessment of autotools, but
that can be explained by the problem that autotools try to solve: building
stuff on any UGLIX system released in the last 20 years.
all these fancy new build systems with shiny features have given up on
this problem, and instead try to solve the simpler problem of getting
stuff to build on _recent_, widely-used systems.
If you would offer a migration path from dmake to plain GNU make and
from tcsh to bash, that might be quite interesting ...
certainly GNU make would be an improvement over dmake.
regarding tcsh: i've just recently build DEV300m41 on a system that did
not have tcsh installed, so that should work already.
A recurring theme in OOo conference presentations and similar material
seems to be that hacking on OOo is hard for newcomers partly because
it is such a complex beast to build. Making it easier could bring in
more contributors.
dmake is not what makes OOo complex.
i disagree with that. it is not that dmake makes OOo complex, it is just
another gratuitous difference that _contributes_ to OOo's complexity.
for CMake, there is at least the KDE project which is large, portable
(there are win32 and os x ports in the works), and (afaik) happy with
their choice. that's already lots of people familiar with CMake and
interested in free software.
Have Fun,
Björn Michaelsen
regards,
michael
--
"C++ ... was one of these things like MS-DOS that nobody took seriously
because who would ever fall for a joke like that?" -- Alan Kay
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
