[Scons-users] Problems with the Java builder

[Tom Tanner (BLOOMBERG/ LONDON)]

An answer would be to look at making scons truss (or use interpose libraries) the command it is executing and parse the outbut for files being created/opened (c.f. http://code.google.com/p/fabricate/)

This gives you your input dependencies and your actual generated targets (not just the ones you thought you generated...)

Per se I think fabricator has all the putfalls of make, except
a) it doesn't use timestamps at all (and because scons tries to do things with timestamps I occasionally have problems)
b) it gets the real dependencies and targets

You still need to do scanning however, because that's the only way you can pick up dependencies on generated files (unless you want to be very clever by checking the failed opens then adding the dependency and retrying the build step)

----- Original Message -----
From: scons-users at scons.org
To: dl9obn at darc.de, scons-users at scons.org
At: Jan 24 2014 12:23:01

On Tue, 2014-01-21 at 21:29 +0100, Dirk Bächle wrote:

Dirk,

[…]

> compiling is probably not really required, but parsing enough of Java's 

> syntax to be able to extract the infos about anonymous classes and such. 

> From this, an Emitter could then mimic the naming schemes of the 

> different Java compilers and spit out the correct list of created class 

> files.


True in principle, but… do we want the hassle?

Before going forward with anything, it would be wise to check out what
Gradle has done to support incremental compilation. I would guess that
they visit the compilation output after compilation and record it.


> > Given that the number of people working with Java who are not using

> > Gradle (or Maven (or Ant)) is tiny, there isn't really that much energy

> > available to see what might be done. Given Scala has SBT, Clojure has

> > Leiningen, and Kotlin and Ceylon nigh on require an IDE, and Gradle can

> > now process C and C++, the need to make SCons JVM language offering work

> > is receding rapidly into the distance.

> >

> This sounds to me, as if you don't plan to work on Java support for 

> SCons anymore...which is totally fine. Has there been any further 

> development or discussion together with JanNijtmans, beyond what's 

> currently available at

> 

>    http://www.scons.org/wiki/JavaStrategy


Well Gant is now basically dead, as is Ant. Gradle and Maven are the
Java tools of choice. SBT has a lot of traction amongst Scala users.
Leiningen for Closure ceylon.build for Ceylon — basically each new
language thinks it has to create a new build framework or be deemed
incapable :-)

Analogously, Go, D, and Rust are creating their own build frameworks
rather than using one of SCons, Waf, CMake, Autotools. The core issues
are:

    Lack of support for an all source at once approach.
    Lack of support for packages.


> and

> 

>    http://www.scons.org/wiki/JavaSupport

> 

> ? If yes, it would be great if we could wrap things up a little, by 

> collecting and documenting the additional findings.


I am now firmly of the view that any effort to progress JVM language
support in SCons is wasted effort. So I think deleting those documents,
or at least marking them "dead", is easier than trying to bring them up
to date.

SCons is fundamentally a C compilation system. It happens to work for C
++ (at least until C++ modules hit the streets) and LaTeX (My main use
these days!). Fortran got grafted in; mostly successfully I gather, I
don't do any Fortran that uses modules. D is a bit of a handful since
for application construction you do an "all at once" build of all
sources. For D the SCons compilation model of source → object is only
really useful for library building.


> I'll continue my work on an Ant-like recursive Glob() soon, which would 

> also help for Java projects. If we then overcome the Emitter problem 

> somehow, do you think we could get back in the game? Or are the other 

> obstacles you mentioned, like support for code coverage tools, equally 

> big and hard to tackle?


I think this would a good thing whatever happens re Java support. I
often want to use this in SCons already :-)

-- 
Russel.
=============================================================================
Dr Russel Winder      t: +44 20 7585 2200   voip: sip:russel.winder at ekiga.net
41 Buckmaster Road    m: +44 7770 465 077   xmpp: russel at winder.org.uk
London SW11 1EN, UK   w: www.russel.org.uk  skype: russel_winder

_______________________________________________
Scons-users mailing list
Scons-users at scons.org
http://four.pairlist.net/mailman/listinfo/scons-users