I spent some time trying out SCons (http://www.scons.org). I think Barry
mentioned it at one
point on this list. Here is a minimal example:
The need right now is to find a way to allow SCons to use the dependency
information generated
by ‘patsopt --depgen’. It should be a little task for someone familiar with
Python. Any volunteers?
I will wait
I can’t say that I disagree; I’m probably worrying too much about pleasing
Java developers that I may or may not need to interact with.On Sun, Dec 7, 2014 at 7:57 PM, gmhwxi gmh...@gmail.com wrote:
For individual projects, I don’t feel that such “big” pictures matter so
much. As long as
you are prepared to “get burnt”, it should be fine to try whatever is out
there.
In terms of dependency handling, you can always hand-code it using your
own scripts.
Yes, it is ad hoc/unpleasant but a few dozens lines of Perl code can go a
long way
On Sunday, December 7, 2014 6:21:25 PM UTC-5, gmhwxi wrote:
I spent some time trying out SCons (http://www.scons.org). I think Barry
mentioned it at one
point on this list. Here is a minimal example:
The need right now is to find a way to allow SCons to use the dependency
information generated
by ‘patsopt --depgen’. It should be a little task for someone familiar
with Python. Any volunteers?
I will wait
The author of the original email (Russel Winder) is clearly very
knowledgeable in this area.
I agree with most of his observations. To me, the question we should ask
is: why would
a person X use SCons to build a Java-based project?
I think the most likely scenario is that the person is already comfortable
with SCons but has
had very little exposure to tools like Ant and Maven. As was suggested in
the essay, SCons
should focus on C/C++/Fortran for now before moving into Java. I feel the
same. Also, the
field is not stable/static. It is a good bet that more build tools for Java
will be out in the near
future.On Mon, Dec 8, 2014 at 6:06 PM, Barry Schwartz < chemoe...@chemoelectric.org> wrote:
One thing about the essay at JavaStrategy · SCons/scons Wiki · GitHub is
it points out exactly what are some of the most noticeable problems
with SCons. It is not afraid to point these out.
I usually describe SCons as making straightforward and easy things
that are very difficult in Make, while making incomprehensibly
difficult (for me) some things that should be, and in Make are,
quite simple.
Probably some of these matters could be addressed by writing one’s own
clever set of modules.
Also, for dependency management, I believe Ant and Maven have their own
solutions. Ant uses Ivy, which is also used by Gradle.On Mon, Dec 8, 2014 at 12:25 AM, gmhwxi gmh...@gmail.com wrote:
Speaking of dependency analysis, what would be a good tool for Java?
On Monday, December 8, 2014 12:02:30 AM UTC-5, gmhwxi wrote:
Also, I think it is healthy to have more than one way to build a system.
I still remember vividly the horror I experienced when using Eclipse to do
a project in Java several years ago. Making a minor change would trigger
Eclipse to very slowly re-compile more than 80 files. I simply could not
figure
out how Eclipse did dependency analysis. Instead, I switched back to emacs
plus Makefiles (plus hand-code dependency).
On Sunday, December 7, 2014 8:54:30 PM UTC-5, Brandon Barker wrote:
I can’t say that I disagree; I’m probably worrying too much about
pleasing Java developers that I may or may not need to interact with.
For individual projects, I don’t feel that such “big” pictures matter
so much. As long as
you are prepared to “get burnt”, it should be fine to try whatever is
out there.
In terms of dependency handling, you can always hand-code it using your
own scripts.
Yes, it is ad hoc/unpleasant but a few dozens lines of Perl code can go
a long way
On Sunday, December 7, 2014 6:21:25 PM UTC-5, gmhwxi wrote:
I spent some time trying out SCons (http://www.scons.org). I think
Barry mentioned it at one
point on this list. Here is a minimal example:
The need right now is to find a way to allow SCons to use the
dependency information generated
by ‘patsopt --depgen’. It should be a little task for someone familiar
with Python. Any volunteers?
I will wait
Also, I think it is healthy to have more than one way to build a system.
I still remember vividly the horror I experienced when using Eclipse to do
a project in Java several years ago. Making a minor change would trigger
Eclipse to very slowly re-compile more than 80 files. I simply could not
figure
out how Eclipse did dependency analysis. Instead, I switched back to emacs
plus Makefiles (plus hand-code dependency).On Sunday, December 7, 2014 8:54:30 PM UTC-5, Brandon Barker wrote:
I can’t say that I disagree; I’m probably worrying too much about pleasing
Java developers that I may or may not need to interact with.
On Sun, Dec 7, 2014 at 7:57 PM, gmhwxi <gmh...@gmail.com <javascript:>> wrote:
For individual projects, I don’t feel that such “big” pictures matter so
much. As long as
you are prepared to “get burnt”, it should be fine to try whatever is out
there.
In terms of dependency handling, you can always hand-code it using your
own scripts.
Yes, it is ad hoc/unpleasant but a few dozens lines of Perl code can go a
long way
On Sunday, December 7, 2014 6:21:25 PM UTC-5, gmhwxi wrote:
I spent some time trying out SCons (http://www.scons.org). I think
Barry mentioned it at one
point on this list. Here is a minimal example:
The need right now is to find a way to allow SCons to use the dependency
information generated
by ‘patsopt --depgen’. It should be a little task for someone familiar
with Python. Any volunteers?
I will wait
Speaking of dependency analysis, what would be a good tool for Java?On Monday, December 8, 2014 12:02:30 AM UTC-5, gmhwxi wrote:
Also, I think it is healthy to have more than one way to build a system.
I still remember vividly the horror I experienced when using Eclipse to do
a project in Java several years ago. Making a minor change would trigger
Eclipse to very slowly re-compile more than 80 files. I simply could not
figure
out how Eclipse did dependency analysis. Instead, I switched back to emacs
plus Makefiles (plus hand-code dependency).
On Sunday, December 7, 2014 8:54:30 PM UTC-5, Brandon Barker wrote:
I can’t say that I disagree; I’m probably worrying too much about
pleasing Java developers that I may or may not need to interact with.
For individual projects, I don’t feel that such “big” pictures matter so
much. As long as
you are prepared to “get burnt”, it should be fine to try whatever is
out there.
In terms of dependency handling, you can always hand-code it using your
own scripts.
Yes, it is ad hoc/unpleasant but a few dozens lines of Perl code can go
a long way
On Sunday, December 7, 2014 6:21:25 PM UTC-5, gmhwxi wrote:
I spent some time trying out SCons (http://www.scons.org). I think
Barry mentioned it at one
point on this list. Here is a minimal example:
The need right now is to find a way to allow SCons to use the
dependency information generated
by ‘patsopt --depgen’. It should be a little task for someone familiar
with Python. Any volunteers?
I will wait
For individual projects, I don’t feel that such “big” pictures matter so
much. As long as
you are prepared to “get burnt”, it should be fine to try whatever is out
there.
In terms of dependency handling, you can always hand-code it using your own
scripts.
Yes, it is ad hoc/unpleasant but a few dozens lines of Perl code can go a
long way :)On Sunday, December 7, 2014 6:21:25 PM UTC-5, gmhwxi wrote:
I spent some time trying out SCons (http://www.scons.org). I think Barry
mentioned it at one
point on this list. Here is a minimal example:
The need right now is to find a way to allow SCons to use the dependency
information generated
by ‘patsopt --depgen’. It should be a little task for someone familiar
with Python. Any volunteers?
I will wait
It should be possible (if one desires) to get SCons to compile an ATS
program with thorough dependency analysis and without calling out to
the shell. With ATS it should probably work nicely using diffs rather
than timestamps, too. (I found that unreliable with OCaml, at least in
my own setup.)