sunstudio

Added by Jonathan Adams over 3 years ago

there are no files in the sunstudio package left in the Oracle repositories, it looks like they've removed access to the compilers for new people.

I doubt that they can force an update on my b134 system, but I've just disabled my repositories in the hopes that they won't try to update and remove my software.

root@jadlaptop:~# pkg contents sunstudio
PATH
root@jadlaptop:~# pkg info sunstudio
Name: developer/sunstudio
Summary:
State: Installed
Publisher: opensolaris.org
Version: 12.1.1
Build Release: 5.11
Branch: 0.111
Packaging Date: March 6, 2010 12:22:34 AM
Size: 0.00 B
FMRI: pkg://opensolaris.org/developer/,5.11-0.111:20100306T002234Z

although I could just be missing something ...


Replies (21)

RE: sunstudio - Added by Jonathan Adams over 3 years ago

http://opensolaris.org/jive/thread.jspa?messageID=495020&tstart=0

"I downloaded Sun Studio 12 using my personal Sun account at home without any problem last Saturday.

Tried to download it at work today and got the following message: "You have selected Content/Patches which requires a Specific type of Contract to Access." I switched to my business account and it worked just fine. "

RE: sunstudio - Added by Bernd Helber over 3 years ago

Hi Guys

same to issue to me, seems that oracle takes it serious?
I'm kinda wondered.

RE: sunstudio - Added by Nigel Smith over 3 years ago

I just checked, and I cannot now download the SunStudio 12 compiler :-(.
It looks like it's no longer a 'free' download - you need a 'contract' with Oracle.
I'm fairly sure I downloaded it earlier this year, but I've searched my disks,
and I cannot currently find it :-(.
I seem to remember that there were problems compiling opensolaris with gcc.
Can anyone give an up-to-date status regarding using gcc?
Thanks
Nigel

RE: sunstudio - Added by Nigel Smith over 3 years ago

Hi, Russ.
No, that link does not work for me.
The download links still end up going to the page saying:
"You have selected Content/Patches which requires a Specific type of Contract to Access."

RE: sunstudio - Added by peter ducai over 3 years ago

but you can still download studio express on http://www.oracle.com/technetwork/server-storage/sunstudio/downloads/index-jsp-136197.html

isn't it enough for compiling?

RE: sunstudio - Added by Jonathan Adams over 3 years ago

Not that I'm a fan of twitter but:

Phil Jaenke says: http://twitter.com/rootwyrm

"Apparently Sun Studio going unavailable appears to have been an unintentional screwup in access control. #Illumos about 13 hours ago via TweetDeck"

If you believe that, that is :)

RE: sunstudio - Added by Nigel Smith over 3 years ago

Hi Jonathan,
Thanks for that twitter tip.

This link:

http://www.oracle.com/technetwork/server-storage/sunstudio/downloads/index.html

..does still says it's a "Free Download".

So lets hope Oracle can fix this 'access control' issue.
Presumably someone has told them about it, and they are working on it...?

The download link for SolarisStudioExpress2010-06-sol-x86.tar.bz2 is still working.

But I was under the impression you needed a 'special' version of Studio,
for the latest builds of OpenSolaris &/or Illumos.

So I'm not sure if this 'Express' version is sufficient.
Please can anyone confirm this?

RE: sunstudio - Added by Eric Lowe over 3 years ago

My two cents, I think you may want to consider standardizing on gcc for your toolchain.

First, if the goal of this project is complete emancipation from encumbered bits [1], it seems counterproductive to rely on the Studio compilers, because now you will still need all developers to go get the "free" compilers before they can compile anything.

Moreover, gcc is a far more advanced and performant compiler than it used to be, and at least for OS development it has a lot of features which are extremely useful that aren't present in the Studio suite, such as inline assembly and true inline functions.

- Eric

[1] Lack of emancipation I always viewed as the biggest stumbling block for OpenSolaris, so I'm 100% behind this project!

RE: sunstudio - Added by Ivan Nudzik over 3 years ago

Ad. performance of gcc: SunStudio doing more speedy code than gcc on SPARC.
In fact it is crucial for Illumos to leave SunStudio dependency to be successful.

RE: sunstudio - Added by Yong Liu over 3 years ago

LLVM could be a viable alternative - it's faster than GCC and the compiled code runs faster. And LLVM is also an open sourced project.

Apple has replaced GCC compiler with LLVM in it's upcoming XCode 4.

RE: sunstudio - Added by Evgeny ILyin over 3 years ago

Studio compilers should be still available for free download. That is for sure..

RE: sunstudio - Added by Evgeny ILyin over 3 years ago

Eric, what version of the gcc are you proposing for this project? Also Studio compilers are still free for use. Also with further development how we can be assured that gcc will still be good compiler for illumos?

Thanks.

Eric Lowe wrote:

My two cents, I think you may want to consider standardizing on gcc for your toolchain.

First, if the goal of this project is complete emancipation from encumbered bits [1], it seems counterproductive to rely on the Studio compilers, because now you will still need all developers to go get the "free" compilers before they can compile anything.

Moreover, gcc is a far more advanced and performant compiler than it used to be, and at least for OS development it has a lot of features which are extremely useful that aren't present in the Studio suite, such as inline assembly and true inline functions.

- Eric

[1] Lack of emancipation I always viewed as the biggest stumbling block for OpenSolaris, so I'm 100% behind this project!

RE: sunstudio - Added by Evgeny ILyin over 3 years ago

Why it is so critical? I've heard recenly that you cannot build ON with new gcc. If it is true, that moving project to the different compiler could make startup time for the project longer than it could be.

Ivan Nudzik wrote:

Ad. performance of gcc: SunStudio doing more speedy code than gcc on SPARC. In fact it is crucial for Illumos to leave SunStudio dependency to be successful.

RE: sunstudio - Added by M. Kuczev over 3 years ago

Please can anyone confirm this?

Yes, I have the same issue here. It says I need a "Specific type of Contract to Access". I have all files that used to be free to download (both as tarfiles and pkgs/rpms), so if Oracle doesn't fix the issue soon, I can put them on public FTP. You may find the links below useful, too.

ftp://ftp.sovintel.ru/pub/software/Solaris/stuff/SunStudio-12u1/
ftp://ftp.nstu.ru/pub/vendors/Sun/SunStudio12u1/

Best regards

RE: sunstudio - Added by Ivan Nudzik over 3 years ago

Evgeny ILyin wrote:

Why it is so critical? I've heard recenly that you cannot build ON with new gcc. If it is true, that moving project to the different compiler could make startup time for the project longer than it could be.

Ivan Nudzik wrote:

Ad. performance of gcc: SunStudio doing more speedy code than gcc on SPARC. In fact it is crucial for Illumos to leave SunStudio dependency to be successful.

I know that ON can't build only with gcc yet. This in fact can compare to balls of Illumos that Oracle hold in hand and any time can smash a bit more if they want. ;-) Or?
Skip of SunStudio dependency is not condition of startup, rather a process that should start in parallel with startup. It is not a trivial task to not to break up merges from OpenSolaris. And for why?
  • ad. paragraph above
  • there is an aim to potentially port to Power, ARM, MIPS64 and those are not targets of SunStudio.
  • I think that there are some legal issues on code produced by SunStudio so it is not only about "can download" vs. "can't download". I'm not 100% sure about it right now...
  • applications - many of OSS apps can only build with gcc and some assembler optimized ones also require more later binutils too. It is easier to build everything with one compiler for those that potentially want adopt Illumos. In fact gcc+binutils combination is horrible to use on *Solaris yet (good headers and linker scripts are needed). Applications rules many operating system movements even in case that Illumos is ON only.
  • In fact SUN made many changes in Solaris headers and SunStudio to be compatible with gcc, to be even able to port valuable amount of OSS apps to OpenSolaris. It is questionable what is better, to rape SunStudio to be compatible with gcc, or has ON build-able with gcc (and skip SunStudio for whole). Illumos should not deal with such a business question as SUN must. It is more simple: gcc build-able => far more wider community, far more applications.
  • Solaris/OpenSolaris hold many too old backward compatibilities and is questionable how much of them has meaning for Illumos, cause Illumos has no long term support deals. I'm not deep in ON sources, but I think that those backward compatibilities are at the first 'fear & loathing' for gcc.
  • BSDs are trying to skip gcc cause of GPL, but they still relaying on it, cause for this kind of community projects nowadays is valid 'gcc is a must and others are valuable'. LLVM gaining focus fast, but now it is not as 'production ready' as gcc is. Optimizer from LLVM shuld be plugged in to gcc, the same way as SUN did it with CoolTools gcc-fss.

That's how I see it...

RE: sunstudio - Added by A Hettinger over 3 years ago

Supporting GCC (or whatever) is fine, but maintaining support for SS is also a must. Remember we aren't trying to through our toys out of the pram here. We want to be able to work Oracle in the future, and saying "we are breaking support for your compiler without a good reason" is going to make sure they can't work with us almost as fast as licensing issues. (note, I don't think this is what you are proposing, but I just want to keep people on the same page)

there is an aim to potentially port to Power, ARM, MIPS64 and those are not targets of SunStudio.

True, and this is probably the single most compelling argument for GCC support I can think of.

I think that there are some legal issues on code produced by SunStudio so it is not only about "can download" vs. "can't download". I'm not 100% sure about it right now...

That's going to require some backing. I do not believe there is any encumbrances with the current version of SS's license with regard to binaries produced.

applications - many of OSS apps can only build with gcc and some assembler optimized ones also require more later binutils too. It is easier to build everything with one compiler for those that potentially want adopt Illumos. In fact gcc+binutils combination is horrible to use on *Solaris yet (good headers and linker scripts are needed). Applications rules many operating system movements even in case that Illumos is ON only.

Most of these are GNUisums that should be squashed whenever found, and are relatively easy to fix anyway. Honestly, the hardest problem I've encountered with porting from GCC to SS is the default libstdc++ in SS is not complete, and not all libraries are build with libstlport. Over on the other side of the fence (in Sun/Oracle land) there was a push to integrate Apache's library, which if we transitioned over to would make this a non-issue.

In fact SUN made many changes in Solaris headers and SunStudio to be compatible with gcc, to be even able to port valuable amount of OSS apps to OpenSolaris. It is questionable what is better, to rape SunStudio to be compatible with gcc, or has ON build-able with gcc (and skip SunStudio for whole). Illumos should not deal with such a business question as SUN must. It is more simple: gcc build-able => far more wider community, far more applications.

SunStudio doesn't need to be compatible with GCC, it just needs to be compatible with the the C and C++ languages, and barring the default libstdc++ it is (stupid templates). Assuming GCC provides an actual C and C++ compiler (not just something vaguely near), the only issue is to handle differences in compiler flags. This has been a solved problem in computer-science for some time, and insofar as various projects have not figured it out, patches should be submitted.

Solaris/OpenSolaris hold many too old backward compatibilities and is questionable how much of them has meaning for Illumos, cause Illumos has no long term support deals. I'm not deep in ON sources, but I think that those backward compatibilities are at the first 'fear & loathing' for gcc.

There is no question, those backward compatibilities are a must. One of our goals is ABI compatibility, that means that if someone builds a binary on Solaris or Indiana, they can copy it over to an Illumos build and it will just run (provided it is the same architecture and there aren't any unmet library dependencies, ofcourse). Our reluctance to change the ABI is one of the (list of) reasons why I prefer Solaris to Linux. It's a Good Thing (tm).

BSDs are trying to skip gcc cause of GPL, but they still relaying on it, cause for this kind of community projects nowadays is valid 'gcc is a must and others are valuable'. LLVM gaining focus fast, but now it is not as 'production ready' as gcc is. Optimizer from LLVM shuld be plugged in to gcc, the same way as SUN did it with CoolTools gcc-fss.

If this is true (I don't follow the BSDs much) it's just generic license trolling on their part. The resulent binaries are not a derrivitive work of the compiler, therefore the GPL would make no restrictions on them. Furthermore, as much as I love the UIUC (I live just slightly north of the Urbana-Champaign campus), I find that integrating their compiler improvements into GCC are slightly out of scope for this project.

RE: sunstudio - Added by Ivan Nudzik over 3 years ago

A Hettinger wrote:

Supporting GCC (or whatever) is fine, but maintaining support for SS is also a must. Remember we aren't trying to through our toys out of the pram here. We want to be able to work Oracle in the future, and saying "we are breaking support for your compiler without a good reason" is going to make sure they can't work with us almost as fast as licensing issues. (note, I don't think this is what you are proposing, but I just want to keep people on the same page)

there is an aim to potentially port to Power, ARM, MIPS64 and those are not targets of SunStudio.

True, and this is probably the single most compelling argument for GCC support I can think of.

After Friday the 13th, do you still think that's the only reason?

I think that there are some legal issues on code produced by SunStudio so it is not only about "can download" vs. "can't download". I'm not 100% sure about it right now...

That's going to require some backing. I do not believe there is any encumbrances with the current version of SS's license with regard to binaries produced.

If I remember right, there are some IP on inline libraries (math & other) and microtasking libraries. These aren't in fact generally used, but what if there is more such cases in code produced by SS?

applications - many of OSS apps can only build with gcc and some assembler optimized ones also require more later binutils too. It is easier to build everything with one compiler for those that potentially want adopt Illumos. In fact gcc+binutils combination is horrible to use on *Solaris yet (good headers and linker scripts are needed). Applications rules many operating system movements even in case that Illumos is ON only.

Most of these are GNUisums that should be squashed whenever found, and are relatively easy to fix anyway. Honestly, the hardest problem I've encountered with porting from GCC to SS is the default libstdc++ in SS is not complete, and not all libraries are build with libstlport. Over on the other side of the fence (in Sun/Oracle land) there was a push to integrate Apache's library, which if we transitioned over to would make this a non-issue.

STL apps can't compile with SS almost in general, but not only those.

In fact SUN made many changes in Solaris headers and SunStudio to be compatible with gcc, to be even able to port valuable amount of OSS apps to OpenSolaris. It is questionable what is better, to rape SunStudio to be compatible with gcc, or has ON build-able with gcc (and skip SunStudio for whole). Illumos should not deal with such a business question as SUN must. It is more simple: gcc build-able => far more wider community, far more applications.

SunStudio doesn't need to be compatible with GCC, it just needs to be compatible with the the C and C++ languages, and barring the default libstdc++ it is (stupid templates). Assuming GCC provides an actual C and C++ compiler (not just something vaguely near), the only issue is to handle differences in compiler flags. This has been a solved problem in computer-science for some time, and insofar as various projects have not figured it out, patches should be submitted.

GCC is a bit vague - not so strict - in C/C++ language standards and also has some own constructs which makes it 'speak own dialect' and this is the 'dialect' of most OSS apps. Rather to know dialect (use GCC), than be dependent on spokesman (code patches), isn't it like that?
I used to compile OSS apps on Solaris with pkgsrc, which employs wrappers for compilers to translate compiler flags, so no problem here. I've seen valuable step in SS ability to compile OSS apps in about last 3y and it is not demand of more patches in pkgsrc of OSS apps to conform standards.

Solaris/OpenSolaris hold many too old backward compatibilities and is questionable how much of them has meaning for Illumos, cause Illumos has no long term support deals. I'm not deep in ON sources, but I think that those backward compatibilities are at the first 'fear & loathing' for gcc.

There is no question, those backward compatibilities are a must. One of our goals is ABI compatibility, that means that if someone builds a binary on Solaris or Indiana, they can copy it over to an Illumos build and it will just run (provided it is the same architecture and there aren't any unmet library dependencies, ofcourse). Our reluctance to change the ABI is one of the (list of) reasons why I prefer Solaris to Linux. It's a Good Thing (tm).

I prefer Solaris for the same reason, but is there really a need for Illumos to be binary compatible with Solaris 8 and older? I mean specialties, that needs versions on symbols in libraries and nowadays are better handled with recent functions.

BSDs are trying to skip gcc cause of GPL, but they still relaying on it, cause for this kind of community projects nowadays is valid 'gcc is a must and others are valuable'. LLVM gaining focus fast, but now it is not as 'production ready' as gcc is. Optimizer from LLVM shuld be plugged in to gcc, the same way as SUN did it with CoolTools gcc-fss.

If this is true (I don't follow the BSDs much) it's just generic license trolling on their part. The resulent binaries are not a derrivitive work of the compiler, therefore the GPL would make no restrictions on them. Furthermore, as much as I love the UIUC (I live just slightly north of the Urbana-Champaign campus), I find that integrating their compiler improvements into GCC are slightly out of scope for this project.

I agree with you, it's only comment on previous posts about LLVM...

Upon statement of Oracle about releasing code to public after binary release, Illumos became a fork of OpenSolaris. Every period between merge of Oracle code it is a fork even if not told loud. Or Illumos will not accept any commit from community at all, only builds from pure Oracle code?

RE: sunstudio - Added by A Hettinger over 3 years ago

Ivan Nudzik wrote:

A Hettinger wrote:

Supporting GCC (or whatever) is fine, but maintaining support for SS is also a must. Remember we aren't trying to through our toys out of the pram here. We want to be able to work Oracle in the future, and saying "we are breaking support for your compiler without a good reason" is going to make sure they can't work with us almost as fast as licensing issues. (note, I don't think this is what you are proposing, but I just want to keep people on the same page)

there is an aim to potentially port to Power, ARM, MIPS64 and those are not targets of SunStudio.

True, and this is probably the single most compelling argument for GCC support I can think of.

After Friday the 13th, do you still think that's the only reason?

Yes, I still maintain that is the most compelling argument (note, that is not anywhere near the same as the only reason). SS binaries are still redistributable, so they should be properly supported.

I think that there are some legal issues on code produced by SunStudio so it is not only about "can download" vs. "can't download". I'm not 100% sure about it right now...

That's going to require some backing. I do not believe there is any encumbrances with the current version of SS's license with regard to binaries produced.

If I remember right, there are some IP on inline libraries (math & other) and microtasking libraries. These aren't in fact generally used, but what if there is more such cases in code produced by SS?

applications - many of OSS apps can only build with gcc and some assembler optimized ones also require more later binutils too. It is easier to build everything with one compiler for those that potentially want adopt Illumos. In fact gcc+binutils combination is horrible to use on *Solaris yet (good headers and linker scripts are needed). Applications rules many operating system movements even in case that Illumos is ON only.

Most of these are GNUisums that should be squashed whenever found, and are relatively easy to fix anyway. Honestly, the hardest problem I've encountered with porting from GCC to SS is the default libstdc++ in SS is not complete, and not all libraries are build with libstlport. Over on the other side of the fence (in Sun/Oracle land) there was a push to integrate Apache's library, which if we transitioned over to would make this a non-issue.

STL apps can't compile with SS almost in general, but not only those.

So far STL apps are the only things I have found that do not compile due to a deviation from spec. Every other problem I have encountered has been that GCC lets things other then C/C++ be called that. If you have some examples of fully proper C/C++ that does not compile (exempting the known STL problems) I'd be welcome to being made aware.

In fact SUN made many changes in Solaris headers and SunStudio to be compatible with gcc, to be even able to port valuable amount of OSS apps to OpenSolaris. It is questionable what is better, to rape SunStudio to be compatible with gcc, or has ON build-able with gcc (and skip SunStudio for whole). Illumos should not deal with such a business question as SUN must. It is more simple: gcc build-able => far more wider community, far more applications.

SunStudio doesn't need to be compatible with GCC, it just needs to be compatible with the the C and C++ languages, and barring the default libstdc++ it is (stupid templates). Assuming GCC provides an actual C and C++ compiler (not just something vaguely near), the only issue is to handle differences in compiler flags. This has been a solved problem in computer-science for some time, and insofar as various projects have not figured it out, patches should be submitted.

GCC is a bit vague - not so strict - in C/C++ language standards and also has some own constructs which makes it 'speak own dialect' and this is the 'dialect' of most OSS apps. Rather to know dialect (use GCC), than be dependent on spokesman (code patches), isn't it like that? I used to compile OSS apps on Solaris with pkgsrc, which employs wrappers for compilers to translate compiler flags, so no problem here. I've seen valuable step in SS ability to compile OSS apps in about last 3y and it is not demand of more patches in pkgsrc of OSS apps to conform standards.

You say "dialect" I say "something other then C/C++." That's the problem with GNU, no desire to conform to standards, and EXACTLY the reason declaring it the "one compiler to rule them all" is wrong.

Solaris/OpenSolaris hold many too old backward compatibilities and is questionable how much of them has meaning for Illumos, cause Illumos has no long term support deals. I'm not deep in ON sources, but I think that those backward compatibilities are at the first 'fear & loathing' for gcc.

There is no question, those backward compatibilities are a must. One of our goals is ABI compatibility, that means that if someone builds a binary on Solaris or Indiana, they can copy it over to an Illumos build and it will just run (provided it is the same architecture and there aren't any unmet library dependencies, ofcourse). Our reluctance to change the ABI is one of the (list of) reasons why I prefer Solaris to Linux. It's a Good Thing (tm).

I prefer Solaris for the same reason, but is there really a need for Illumos to be binary compatible with Solaris 8 and older? I mean specialties, that needs versions on symbols in libraries and nowadays are better handled with recent functions.

I still see Sol 8 bins floating around, and it is beneficial to maintain compaitibility with Sol 10 (and 11 when it is out). Either their bins run on us or not. I, for one, want them to. As long as Sol 11/10 remain compatible, we have to as well, otherwise we are not compatible with them.

BSDs are trying to skip gcc cause of GPL, but they still relaying on it, cause for this kind of community projects nowadays is valid 'gcc is a must and others are valuable'. LLVM gaining focus fast, but now it is not as 'production ready' as gcc is. Optimizer from LLVM shuld be plugged in to gcc, the same way as SUN did it with CoolTools gcc-fss.

If this is true (I don't follow the BSDs much) it's just generic license trolling on their part. The resulent binaries are not a derrivitive work of the compiler, therefore the GPL would make no restrictions on them. Furthermore, as much as I love the UIUC (I live just slightly north of the Urbana-Champaign campus), I find that integrating their compiler improvements into GCC are slightly out of scope for this project.

I agree with you, it's only comment on previous posts about LLVM...

Upon statement of Oracle about releasing code to public after binary release, Illumos became a fork of OpenSolaris. Every period between merge of Oracle code it is a fork even if not told loud. Or Illumos will not accept any commit from community at all, only builds from pure Oracle code?

Yes they might drift away then we will have to integrate the changes to our code. It's unfortunate that they don't seem to want to work together, but that's to be expected. Maybe we can forge a working relationship at some point in the future, but right now we work with what we have. Also recall that we are claiming to be more of a core development then an actual distro. If one of the down-stream distros wants to deviate, that's fine, but we need to be able to support those who don't.

RE: sunstudio - Added by Nigel Smith over 3 years ago

Just to get back to one of the original topics on this thread...

The 'free' download of SunStudio is now working again, for me.
I tried it yesterday, and had no problems with my 'contract-less' account.

Following the advice here:

http://insanum.com/blog/2010/06/08/how-to-build-opensolaris

..that you will need two compiler packages.:
SunStudio 12 for compiling and SunStudio 12u1 for lint.
I used links from these pages:

http://hub.opensolaris.org/bin/view/Community+Group+tools/sun_studio_tools
http://hub.opensolaris.org/bin/view/Community+Group+tools/sun_studio_12_tools
http://hub.opensolaris.org/bin/view/Community+Group+tools/sun_studio_12u1_tools

and downloaded files:

sunstudio12-patched-ii-2009Sep-sol-x86.tar.bz2
sunstudio12u1-patched-ii-2010Feb-sol-x86.tar.gz

So I'm hoping that these will be the ones I need to build Illumos...
Regards
Nigel Smith

RE: sunstudio - Added by Hillel Lubman over 3 years ago

There was some announcement before: http://www.opensolaris.org/jive/thread.jspa?messageID=497301
that PathScale can release their optimized C++ compilers (based on Open64) for OpenSolaris x86_64. (But I'm not sure about SPARC). This might be helpful.

(1-21/21)