Re: [Sawfish] lack of contributors

[ Thread Index | Date Index | More Archives ]

Thanks for the reply, I felt like I might have been too harsh after
this mail so made the effort to build sawfish from source,


On Thu, Apr 5, 2012 at 4:33 AM, Christopher Roy Bratusek
<nano@xxxxxxxxxxxxx> wrote:
> On Wednesday 04 April 2012 23:35:21 Campbell Barton wrote:
>> Greetz
>> I work as a blender dev and have been involved in encouraging more
>> devs to get involved with our project (with some success? :) ).
>> heres my 2c.
>> *** Make building as EASY AS POSSIBLE ***
>> I might not put this first, except that previously I had a lot of
>> trouble building sawfish, and not just on once off - different PC's,
>> distros, before the project revived too.
>> Probably half of the troubles I have are that I like to install
>> sawfish to a custom prefix /opt/sawfish for eg. This means I have to
>> point to a custom librep, librep/gtk etc... which i couldn't always
>> get working.
>> Recently I tried building with Aur on arch and had a very strange
>> error running `make install`. where I had to delete a file after make
>> started but before it finished else it would complain the file was
>> already existing and quit. - Not sure if this is Aur-pkg's error or
>> sawfish.
>> While I cant tell you how to run things, it would be really nice if
>> there was some way to checkout librep/rep-gtk/sawfish at once and
>> build all at once.
>> Possibly this could be done with git-submodules?
>> The root dir could have some shell script to build all (so this is
>> 10min work, not a large restructure).
>> As an example - heres a shell script to download/build and install
>> llvm/clang -
>> Correct me if Im wrong but last I tried, sawfish was more trouble to
>> install then this (if a window manager is harder to build then a C/C++
>> compiler.... somethings wrong?)
>> also note that I've built openbox, fluxbox, icewm, notion,
>> windowmaker, jwm - and sawfish is by far the most complicated to get
>> built and running.
>> (on a side note I used to build my own kernel. gcc. glibc xorg, Im
>> familier with building different software and found sawfish quite
>> difficult to build)
>> --- take this with a grain of salt, perhaps I missed something here or
>> just had a lot of bad luck
> I've spend lots of time improving the build-stuff. It's way better than the
> original stuff Sawfish had earlier. For installing into a different prefix
> all you need to to additionally do is:
> export PKG_CONFIG_PATH="$PKG_CONFIG_PATH:/op/sawfish/lib/pkgconfig"
> Or whereever you installed Sawfish.
> Personally I'm not a fan of all-in-one repos. You can still checkout all 5
> separately and use a script in the toplevel-dir to to stuff.

I can understand why you don't like all-in-one repos, so perhaps this
is just more of a documentation issue - basically,
if you choose to have your source configuration that requires more
steps, then these should be well documented, made into a script or
somehow made so new developers don't have to spend too much time
reading ./configure --help and troubleshooting what should be a
straightforward process.

>> For me. not being able to do this....
>> git pull ****; ./configure --prefix=/opt/sawfish; make install
>> ... is a big stumbling block for getting involved with the source ---
>> (or svn/hg/bzr, - scons/cmake etc equivalents)
> Basically no project that uses autotools lets you do that. All require
> in first place, but well, if one is bored he could port Sawfish
> to CMake ...

For this, I only meant that standard commends are used. I didnt mean
to highlight the differences between running ./configure /,
ofcourse this is fine.
For sawfish standard commands ARE used, its only installing into
custom locations thats not easy to find in the docs.

Since i know to look for it now - PGK_CONFIG_PATH is mentioned, though
there could be a section "Installing to a User Defined Path", rather
then documenting this as some exceptional case if you get an error
with one of the libraries.

IMHO this should be added to docs since the docs already state:

"""Caution: This is a raw installation into /usr/local without package

Rather then suggesting they use a package instead, (if they are
interested in building from source they could be interested
developers), better document installing to a custom location.

> Debian is the most tested distribution to build Sawfish on. It works just
> perfect here. According to Teika and Fuchur Gentoo is fine, too. Kim has
> provided improved spec-files to aid building on Fedora.

This is comparing casual developers who like to build their own
versions with package maintainers who have enough time to setup spec,
read other packages spec files etc. When I build/install - LLVM's SVN
for instance Im happy that its not all that complicated and much less
work then writing a spec file and uses standard commands to install
(for CMake in this case).

Developers tend to work on svn/HEAD git/MASTER... so having working
spec files is not so relevant if someone wants to grab the source and
build it IMHO.

> On those distros a simple configure/make/install triplette is enough in the
> outermost cases. If you ran into issues ... you may report them to us.
> Other distributions are somewhat untested/less tested. Since we are provided
> with patches for OpenBSD as soon as something is broken, it should be fine,
> too. FreeBSD and NetBSD also provide it in their ports.
> Regards

- Campbell

Sawfish ML

Mail converted by MHonArc 2.6.19+