9 ms·

The Deletion of gcj

284 points by mynameislegion 10y ago

myhf 10y ago
Google cache: http://webcache.googleusercontent.com/search?q=cache:zFK8UNRnBpkJ:tromey.com/blog/%3Fp%3D911 http://webcache.googleusercontent.com/search?q=cache:zFK8UNR...
okket 10y ago
GCJ was deemed obsolete in 2010, amazing that it was still around until now... http://stackoverflow.com/questions/3032727/java-jre-vs-gcj http://stackoverflow.com/questions/3032727/java-jre-vs-gcj Did anyone used GCJ for something in production?
- marktangotango 10y ago
  I evaluated gcj to use in a shareware desktop application in the 2000's. The goal was to compile java + SWT into native binaries for distribution. The goal was a slim installer that didn't require the user to futz with installing java, or bunding a jre. Ended up going with Excelsior JET which is a fantastic project. In the end, the full java spec doesn't lend itself to aot, because if you claim to be "real" java you have to include support for runtime classloading, which means there has to be a bytecode interpreter embedded in there somewhere. My thoughts are now, if you really want a java/c# style aot language, vala is the way forward.
  - amyjess 10y ago
    > My thoughts are now, if you really want a java/c# style aot language, vala is the way forward. What about D?
  - my123 10y ago
    C# fully supports AOT since a while.
    - swift 10y ago
      I'm pretty sure that C# AOT does not eliminate the need for a JIT in the runtime system. Just as one example, how could expression trees be implemented without one?
      - pjmlp 10y ago
        You cannot use them. In .NET Native you have to explicitly inform the compiler of any kind of reflection behavior you might require.
        zgramana 10y ago
        Not quite right. Mono has the lovely mkbundle tool which can do what the poster wants: AOT your intermediate/byte code and build a statically linked executable containing your app, the runtime, and all your deps (except for key system dynamic libraries like libc or Security.framework, similar to how go works). No restrictions on reflection, unlike .NET Native (which has slightly different design goals). I've used it many times for just this reason with great results. Plus, the runtime was re-licensed to MIT/X11.
        pjmlp 10y ago
        My .NET experience is only on Windows, thanks for the clarification.
- krylon 10y ago
  IIRC, Fedora at one point shipped some Java programs compiled with GCJ, for example Eclipse.
- lmm 10y ago
  I was part of a student society that used GCJ for the code that ran everything (including generating the website). I have not-so-fond memories of the porting effort to get it working on Sun Java and MySQL 5 (it had originally used 4). It was a different time to be sure.
- int_19h 10y ago
  One nice thing that GCJ had was fairly smooth integration with C++ via CNI - much higher level than JNI, and easier to use. https://gcc.gnu.org/onlinedocs/gcj/About-CNI.html#About-CNI https://gcc.gnu.org/onlinedocs/gcj/About-CNI.html#About-CNI Could be used for scripting and extensibility purposes.
dragonquest 10y ago
This is why you should always create something, even if you think it will never be accepted by a large audience. It is worth it for the journey itself. Reading this post made me wish I had spent more time with 'gcj'. Though in all honesty I know I gave up on it because I didn't want to go the tougher way - the the normal JDK was just there. Creation is beautiful, and its ending bittersweet for the creator and the witnesses.
- TheCondor 10y ago
  I live in the moment! Enjoy and cherish those magical times when the work is right, the team is right, and you're having fun. Don't let the stupid stuff bother you. You can rebuild teams like that, find those projects, etc.. but it feels like an eternity when you've had it and then don't have it for a while.
rekado 10y ago
This is a little sad because this means we will have to keep around old versions of GCC with GCJ in order to be able to bootstrap the JDK from source (without a binary bootstrap JDK). Only OpenJDK6 (with the extensive IcedTea patches) can be bootstrapped with GCJ alone, so we also have to keep OpenJDK6 around in order to be able to build the latest JDK. Bootstrapping becomes harder and more legacy software needs to be accumulated in order to actually build new software from source. (I'm working on GNU Guix, a functional package manager, and I wrote the package definitions for the JDKs.)
- josteink 10y ago
  As someone who worked on porting Microsoft .NET Core to new platforms, that's definitely a pain I felt. At first, very few assumptions was made about the Unix bootstrapping process, and the mantra was that you should build the managed components on a Windows machine which already had .NET. Just to get something going, the Unix community decided to use mono to bootstrap the process on non-Windows platforms. This effectively meant you could bootstrap .NET Core on almost any platform where mono was around. That gave quite some platform reach out of the box, and on overall an improvement much welcomed by Microsoft. But later, after Unixes like Linux and OSX become better supported, the bootstrapping method changed to downloading a previous build of .NET Core for the current platform and using that to build the latest sources. I.e. going self-bootstrapping, which sounds fancy on paper. The cost of that? Suddenly it just wasn't doable (without considerable knowledge and effort) to bootstrap .NET Core from source any more on platforms not currently considered first class. I'm not just talking about "strange" outlier OSes like FreeBSD, NetBSD or Alpine Linux or whatever. Just trying to build for Fedora 24 (which doesn't have published packages) fails. This is because the Fedora 23 packages are linked against library-versions which are no longer in Fedora 24. You'd think that be easy to fix, but now, even with Fedora 25 betas starting to ship, there's still no official Fedora 24 package exists from Microsoft, so Fedora users (and developers, and porters, etc) are SOL. So now basically just trying to bootstrap for a previously supported mainstream Linux-distro is seriously PITA, next to impossible. In my honest opinion that's a major regression just for the sake of being self-bootstrapping. So yeah. You should never underestimate the value of having secondary platforms available for bootstrapping your own. You may come to depend on it.
aardvark179 10y ago
http://openjdk.java.net/jeps/8166089 http://openjdk.java.net/jeps/8166089 One Java AOT option goes away, and a new one is arriving.
- bcg1 10y ago
  There is Avian too https://readytalk.github.io/avian/ https://readytalk.github.io/avian/
- iso-8859-1 10y ago
  Hopefully the new one can replace RoboVM as well...
  - chrisseaton 10y ago
    I'm afraid the job of the new AOT is not to produce standalone binaries - it's code to be loaded into the JVM, not to run without it.
lsd5you 10y ago
It was used for creating the native builds of XWT [1] 15 years ago. At the time I found it amazing to see Java programs becoming native (a lot snappier and generally faster too). It seemed (naively) to be beyond what was possible. However, the main problem with GCJ was it used conservative garbage collection - boehm GC - and as such was fundamentally broken for long running programs (what java is good for). Without wanting to denigrate it too much, it seems to me that a lot of 'good' work has been put into this approach which was essentially a cul-de-sac. I had a related rather general thought recently about documenting features in projects. Sometimes the hardest thing to know is whether something is in fact a good idea or not (both in theory and in practice) and the amount of effort/quality of documentation is probably not a good indicator of this. So in the case of the GCJ garbage collector, there was always a lot of talk about how clever the garbage collector was and all the technical aspects, but unless you were already an expert, how would you know if it was infact a good idea or not. [1] That is this one, there are several now doing similar things (!). http://www.xwt.org/ http://www.xwt.org/ The code lives on here https://sourceforge.net/projects/vexi/ https://sourceforge.net/projects/vexi/
- correktlogic 10y ago
  Anyone know if gcj ever experimented with a precise GC? Or did the design of GCC simply not allow for a non-conservative GC implementation?
  - gumby 10y ago
    Historical. We implemented the Bohm GC at Cygnus, funded by Metaphor back around 92/93 if I remember. GCC is / was very conservative.
mclovingoat1 10y ago
I like to fuck sheeps and goats. I love to fuck goat asses and lick sheeps pussies. I like to fuck them all day long. I like to fuck them with gcj – goatfucking_compiler_jerkoff goatfucking_compiler_jerkoff
mclovingoat1 10y ago
I like to fuck sheeps and goats. I love to fuck goat asses and lick sheeps pussies. I like to fuck them all day long. I like to fuck them with gcj – goatfucking_compiler_jerkoff goatfucking_compiler_jerkoff
wodencafe 10y ago
It seems the most beloved feature of GCJ was its ability to compile Java to native code. Is this something that could be implemented in the Java JDK / OpenJDK? Understandably a lot of work, but Oracle is looking at AOT now, for performance reasons. The performance improvements could be taken even further by something like this, Java to native code compilation.
correktlogic 10y ago
What a shame about gcj being dropped. I remember using GCJ with CNI to interface with native C++ code was a much more pleasant experience than using JNI. https://gcc.gnu.org/onlinedocs/gcj/About-CNI.html#About-CNI https://gcc.gnu.org/onlinedocs/gcj/About-CNI.html#About-CNI
fake-name 10y ago
I skimmed the article, and all the comments here, and I still have no fucking idea what "gcj" is. It's apparently related to java. Maybe at least define your acronym once?
- tombert 10y ago
  Yeah, I mean, it's just incredibly hard to go on Wikipedia and look up "gcj". All I got was this extremely informative page about an open source java compiler: https://en.wikipedia.org/wiki/GNU_Compiler_for_Java https://en.wikipedia.org/wiki/GNU_Compiler_for_Java.
akx 10y ago
gcj being killed means someone needs to reimplement pdftk in... well, something else. D:
- LukeShu 10y ago
  Indeed, that was my concern as well. The best solution is probably to replace the C++ bits with Java, and go full JVM.
  - akx 10y ago
    Yep. I recently had to get pdftk running on an old Centos-or-whatever system, and that ended up being a big fat no-go, so I reimplemented the functionality we needed from pdftk (filling forms) with just Java: https://github.com/akx/foofdf https://github.com/akx/foofdf
pritambarhate 10y ago
Not directly related to topic, has anybody used Excelsior JET to compile Java apps to native code? How was your experience? Did it improve the start up time of the app? Did you find it any faster than using a JVM?

284 points by mynameislegion 10y ago

myhf 10y ago

Google cache: http://webcache.googleusercontent.com/search?q=cache:zFK8UNRnBpkJ:tromey.com/blog/%3Fp%3D911 http://webcache.googleusercontent.com/search?q=cache:zFK8UNR...

okket 10y ago

GCJ was deemed obsolete in 2010, amazing that it was still around until now... http://stackoverflow.com/questions/3032727/java-jre-vs-gcj http://stackoverflow.com/questions/3032727/java-jre-vs-gcj Did anyone used GCJ for something in production?

marktangotango 10y ago
I evaluated gcj to use in a shareware desktop application in the 2000's. The goal was to compile java + SWT into native binaries for distribution. The goal was a slim installer that didn't require the user to futz with installing java, or bunding a jre. Ended up going with Excelsior JET which is a fantastic project. In the end, the full java spec doesn't lend itself to aot, because if you claim to be "real" java you have to include support for runtime classloading, which means there has to be a bytecode interpreter embedded in there somewhere. My thoughts are now, if you really want a java/c# style aot language, vala is the way forward.
- amyjess 10y ago
  > My thoughts are now, if you really want a java/c# style aot language, vala is the way forward. What about D?
- my123 10y ago
  C# fully supports AOT since a while.
  - swift 10y ago
    I'm pretty sure that C# AOT does not eliminate the need for a JIT in the runtime system. Just as one example, how could expression trees be implemented without one?
    - pjmlp 10y ago
      You cannot use them. In .NET Native you have to explicitly inform the compiler of any kind of reflection behavior you might require.
      - zgramana 10y ago
        Not quite right. Mono has the lovely mkbundle tool which can do what the poster wants: AOT your intermediate/byte code and build a statically linked executable containing your app, the runtime, and all your deps (except for key system dynamic libraries like libc or Security.framework, similar to how go works). No restrictions on reflection, unlike .NET Native (which has slightly different design goals). I've used it many times for just this reason with great results. Plus, the runtime was re-licensed to MIT/X11.
        pjmlp 10y ago
        My .NET experience is only on Windows, thanks for the clarification.
krylon 10y ago
IIRC, Fedora at one point shipped some Java programs compiled with GCJ, for example Eclipse.
lmm 10y ago
I was part of a student society that used GCJ for the code that ran everything (including generating the website). I have not-so-fond memories of the porting effort to get it working on Sun Java and MySQL 5 (it had originally used 4). It was a different time to be sure.
int_19h 10y ago
One nice thing that GCJ had was fairly smooth integration with C++ via CNI - much higher level than JNI, and easier to use. https://gcc.gnu.org/onlinedocs/gcj/About-CNI.html#About-CNI https://gcc.gnu.org/onlinedocs/gcj/About-CNI.html#About-CNI Could be used for scripting and extensibility purposes.

dragonquest 10y ago

This is why you should always create something, even if you think it will never be accepted by a large audience. It is worth it for the journey itself. Reading this post made me wish I had spent more time with 'gcj'. Though in all honesty I know I gave up on it because I didn't want to go the tougher way - the the normal JDK was just there. Creation is beautiful, and its ending bittersweet for the creator and the witnesses.

TheCondor 10y ago
I live in the moment! Enjoy and cherish those magical times when the work is right, the team is right, and you're having fun. Don't let the stupid stuff bother you. You can rebuild teams like that, find those projects, etc.. but it feels like an eternity when you've had it and then don't have it for a while.

rekado 10y ago

This is a little sad because this means we will have to keep around old versions of GCC with GCJ in order to be able to bootstrap the JDK from source (without a binary bootstrap JDK). Only OpenJDK6 (with the extensive IcedTea patches) can be bootstrapped with GCJ alone, so we also have to keep OpenJDK6 around in order to be able to build the latest JDK. Bootstrapping becomes harder and more legacy software needs to be accumulated in order to actually build new software from source. (I'm working on GNU Guix, a functional package manager, and I wrote the package definitions for the JDKs.)

josteink 10y ago
As someone who worked on porting Microsoft .NET Core to new platforms, that's definitely a pain I felt. At first, very few assumptions was made about the Unix bootstrapping process, and the mantra was that you should build the managed components on a Windows machine which already had .NET. Just to get something going, the Unix community decided to use mono to bootstrap the process on non-Windows platforms. This effectively meant you could bootstrap .NET Core on almost any platform where mono was around. That gave quite some platform reach out of the box, and on overall an improvement much welcomed by Microsoft. But later, after Unixes like Linux and OSX become better supported, the bootstrapping method changed to downloading a previous build of .NET Core for the current platform and using that to build the latest sources. I.e. going self-bootstrapping, which sounds fancy on paper. The cost of that? Suddenly it just wasn't doable (without considerable knowledge and effort) to bootstrap .NET Core from source any more on platforms not currently considered first class. I'm not just talking about "strange" outlier OSes like FreeBSD, NetBSD or Alpine Linux or whatever. Just trying to build for Fedora 24 (which doesn't have published packages) fails. This is because the Fedora 23 packages are linked against library-versions which are no longer in Fedora 24. You'd think that be easy to fix, but now, even with Fedora 25 betas starting to ship, there's still no official Fedora 24 package exists from Microsoft, so Fedora users (and developers, and porters, etc) are SOL. So now basically just trying to bootstrap for a previously supported mainstream Linux-distro is seriously PITA, next to impossible. In my honest opinion that's a major regression just for the sake of being self-bootstrapping. So yeah. You should never underestimate the value of having secondary platforms available for bootstrapping your own. You may come to depend on it.

aardvark179 10y ago

http://openjdk.java.net/jeps/8166089 http://openjdk.java.net/jeps/8166089 One Java AOT option goes away, and a new one is arriving.

bcg1 10y ago
There is Avian too https://readytalk.github.io/avian/ https://readytalk.github.io/avian/
iso-8859-1 10y ago
Hopefully the new one can replace RoboVM as well...
- chrisseaton 10y ago
  I'm afraid the job of the new AOT is not to produce standalone binaries - it's code to be loaded into the JVM, not to run without it.

lsd5you 10y ago

It was used for creating the native builds of XWT [1] 15 years ago. At the time I found it amazing to see Java programs becoming native (a lot snappier and generally faster too). It seemed (naively) to be beyond what was possible. However, the main problem with GCJ was it used conservative garbage collection - boehm GC - and as such was fundamentally broken for long running programs (what java is good for). Without wanting to denigrate it too much, it seems to me that a lot of 'good' work has been put into this approach which was essentially a cul-de-sac. I had a related rather general thought recently about documenting features in projects. Sometimes the hardest thing to know is whether something is in fact a good idea or not (both in theory and in practice) and the amount of effort/quality of documentation is probably not a good indicator of this. So in the case of the GCJ garbage collector, there was always a lot of talk about how clever the garbage collector was and all the technical aspects, but unless you were already an expert, how would you know if it was infact a good idea or not. [1] That is this one, there are several now doing similar things (!). http://www.xwt.org/ http://www.xwt.org/ The code lives on here https://sourceforge.net/projects/vexi/ https://sourceforge.net/projects/vexi/

correktlogic 10y ago
Anyone know if gcj ever experimented with a precise GC? Or did the design of GCC simply not allow for a non-conservative GC implementation?
- gumby 10y ago
  Historical. We implemented the Bohm GC at Cygnus, funded by Metaphor back around 92/93 if I remember. GCC is / was very conservative.

mclovingoat1 10y ago

I like to fuck sheeps and goats. I love to fuck goat asses and lick sheeps pussies. I like to fuck them all day long. I like to fuck them with gcj – goatfucking_compiler_jerkoff goatfucking_compiler_jerkoff

mclovingoat1 10y ago

wodencafe 10y ago

It seems the most beloved feature of GCJ was its ability to compile Java to native code. Is this something that could be implemented in the Java JDK / OpenJDK? Understandably a lot of work, but Oracle is looking at AOT now, for performance reasons. The performance improvements could be taken even further by something like this, Java to native code compilation.

correktlogic 10y ago

What a shame about gcj being dropped. I remember using GCJ with CNI to interface with native C++ code was a much more pleasant experience than using JNI. https://gcc.gnu.org/onlinedocs/gcj/About-CNI.html#About-CNI https://gcc.gnu.org/onlinedocs/gcj/About-CNI.html#About-CNI

fake-name 10y ago

I skimmed the article, and all the comments here, and I still have no fucking idea what "gcj" is. It's apparently related to java. Maybe at least define your acronym once?

tombert 10y ago
Yeah, I mean, it's just incredibly hard to go on Wikipedia and look up "gcj". All I got was this extremely informative page about an open source java compiler: https://en.wikipedia.org/wiki/GNU_Compiler_for_Java https://en.wikipedia.org/wiki/GNU_Compiler_for_Java.

akx 10y ago

gcj being killed means someone needs to reimplement pdftk in... well, something else. D:

LukeShu 10y ago
Indeed, that was my concern as well. The best solution is probably to replace the C++ bits with Java, and go full JVM.
- akx 10y ago
  Yep. I recently had to get pdftk running on an old Centos-or-whatever system, and that ended up being a big fat no-go, so I reimplemented the functionality we needed from pdftk (filling forms) with just Java: https://github.com/akx/foofdf https://github.com/akx/foofdf

pritambarhate 10y ago

Not directly related to topic, has anybody used Excelsior JET to compile Java apps to native code? How was your experience? Did it improve the start up time of the app? Did you find it any faster than using a JVM?