Re: [osol-discuss] Solaris Next timescale
> joerg.schill...@fokus.fraunhofer.de wrote: > > Alan Coopersmith > wrote: > > > >>> CPIO is limited to 8 GB files and the (non-POSIX > ... SVR4) CPIO archive format > >>> Solaris is currently using by default is even > limited to 4 GB file size. > >> So? If the OS installer is unpacking any files > larger than 4gb, there is > >> something seriously wrong. Are there any packages in Solaris 10 or Next that are even 1GB in size? I somehow doubt it. But that wasn't the issue. The issue was, will CPIO or IPS' new on-disk format be used and that's pretty much been answered further down this thread. > > >>> It has been shown already that pkgadd can be > enhanced to use other > >>> archive types and I am strongly against > introducing something new > >>> that is based n the outdated (since y-2001) CPIO. > >> Using cpio in the installs is hardly new. > > > > I was replying to a person who proposed to use cpio > for something new. > > No, he was simply referring to the OS install cd's, > which have used cpio > for a long time. Precisely. There is nothing new about using CPIO for Solaris install CDs or DVDs. Something new would be using an IPS on-disk format for these. ;-) Cheers Andrew. -- This message posted from opensolaris.org ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
> On 07/19/10 03:53 AM, andrew wrote: > ... > > So the big question is: will Solaris Next use IPS > packaging? My money is on IPS packaging but a big fat > CPIO-based DVD installation similar to S10 but using > the new Caiman installers from OpenSolaris (local > text and graphics-based installs plus via serial > line, network and service processor consoles). > > If you've been watching the pkg-discuss mailing list, > you'd note that > the on-disk form for the image packaging system is > already under > development. > > And for the record, the chosen format is not CPIO. Cool. I won't ask you to confirm plans for the Solaris Next DVD install format since that's an Oracle *product*. ;-) Andrew. -- This message posted from opensolaris.org ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
On 07/19/10 03:53 AM, andrew wrote: ... So the big question is: will Solaris Next use IPS packaging? My money is on IPS packaging but a big fat CPIO-based DVD installation similar to S10 but using the new Caiman installers from OpenSolaris (local text and graphics-based installs plus via serial line, network and service processor consoles). If you've been watching the pkg-discuss mailing list, you'd note that the on-disk form for the image packaging system is already under development. And for the record, the chosen format is not CPIO. Cheers, -Shawn ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
--- On Mon, 7/19/10, Dennis Clarke wrote: > > Dennis Clarke wrote: > >> Getting to b144 may be a challenge as I do not > know what patch levels > >> are required or even what compiler to use. > > > Dennis Based on past info using SS 12/GCC: 1. Xorg 7.5: PASS 2. Binutils 2.20.1/GCC 4.5.0: PASS 3. ON snv_144 from ON_142: PASS 4. Coreutils 8.5: PASS 5. BASH 4.1.7: PASS Now that I reconfirmed the use of Sun Studio 12 for the process. Sun Studio 12.1 was not certified for current ON (i.e. snv_144) builds when I checked this awhile ago. I think this is very possible within a 72-hour period based on building the FOSS packages separately beforehand with GCC 4.x. Will try to post some screenshots this weekend (depends!). ~ Ken Mays ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
On 07/19/10 09:21 AM, Alan Coopersmith wrote: I know some work was done on ON to fix gcc 4.x build issues so that the Parfait static analyzer from Sun Labs could be run on the code, but don't know of anyone testing a full build with actual gcc 4.x compilers. I integrated those syntax fixes, in collaboration with the Sun Labs engineers. I haven't tested this for the past few builds, so it's possible that a new syntax problem has crept in. However, based on the last time I tried it, you should be able to build ON for x86 with gcc 4.x. Building for sparc requires syntax fixes in a handful of additional files I haven't integrated yet. Note that I'm talking about syntactic correctness only; I haven't tried to run the resulting binaries. Scott -- Scott Rotondo Senior Principal Engineer, Solaris Core OS Engineering President, Trusted Computing Group Phone: +1 650 786 6309 (Internal x86309) ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
ken mays wrote: > See for ref: > http://hub.opensolaris.org/bin/view/Community+Group+tools/build_instr > > We were using GCC 3.x to compile ON snv_144 and the X consolidation. Still > considered a the preferred method for the community members? I don't think it's ever been preferred, just an option for those who could not or did not want to use the Studio compilers. It won't match the binaries we build, test, and release. > Also, I did remember a note about using GCC 4.x to build ON and the X > consolidation as well in the future. Is this tested by anyone at Oracle > Engineering? We do not test building X with gcc 4.x - only the gcc 3.4.3 from SFW, and that only for the portions that cannot be built with Sun Studio. I would imagine X can be built with gcc 4.x due to upstream work, but have never tested it. I know some work was done on ON to fix gcc 4.x build issues so that the Parfait static analyzer from Sun Labs could be run on the code, but don't know of anyone testing a full build with actual gcc 4.x compilers. -- -Alan Coopersmith-alan.coopersm...@oracle.com Oracle Solaris Platform Engineering: X Window System ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
--- On Mon, 7/19/10, Alan Coopersmith wrote: > From: Alan Coopersmith > Subject: Re: [osol-discuss] Solaris Next timescale > To: dcla...@blastwave.org > Cc: opensolaris-discuss@opensolaris.org > Date: Monday, July 19, 2010, 11:17 AM > Dennis Clarke wrote: > > Getting to b144 may be a challenge as I do not know > what patch levels are > > required or even what compiler to use. > > As always, it's documented on http://hub.opensolaris.org/bin/view/downloads/on > > > I will try with Sun Studio 12.1 > > with all latest patches. I have no idea why people use > 12.1 when that > > compiler is unable to build simple things like GNU > make or libgcrypt > > without blowing up on its own intermediate assembly[1] > output but it seems > > able to build ON. My disdain for Studio 12.1 not > withstanding it does seem > > to be the compiler to use. > > Not yet for building the OS. Studio 12 is > still required - we're testing > a new patchset of Studio 12.1 now for building the OS and > hope to be able > to switch to it soon, but until then, the official build > environment is still > Studio 12 (except that ON uses lint, and only lint, from > 12u1). > > http://hub.opensolaris.org/bin/view/Community+Group+tools/sun_studio_12_tools > has the set to use for now. > > -- > -Alan Coopersmith- > alan.coopersm...@oracle.com > Oracle Solaris Platform > Engineering: X Window System Hi Alan, Please correct me. See for ref: http://hub.opensolaris.org/bin/view/Community+Group+tools/build_instr We were using GCC 3.x to compile ON snv_144 and the X consolidation. Still considered a the preferred method for the community members? Also, I did remember a note about using GCC 4.x to build ON and the X consolidation as well in the future. Is this tested by anyone at Oracle Engineering? ~ Ken Mays ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
> Dennis Clarke wrote: >> Getting to b144 may be a challenge as I do not know what patch levels >> are required or even what compiler to use. > > As always, it's documented on > http://hub.opensolaris.org/bin/view/downloads/on That page is wrong. Or at least not entirely correct. I built this : $ uname -a SunOS aequitas 5.11 snv_138 i86pc i386 i86pc with this : $ cc -V cc: Sun C 5.10 SunOS_i386 Patch 142363-05 2010/04/28 usage: cc [ options] files. Use 'cc -flags' for details The nightly report was clean. I went through this with various people and exchanged emails about it and I'm tired of talking about it. The docs on the Oracle sites are wrong or slightly wrong or slightly correct. I'm not going down that path again .. I'm tired of it and I'm not going to file any more bug reports. I'm sure that Jörg and I can figure it out and who knows, maybe even come up with something that people can use with docs on a site that is community based. All the Xorg bits will come after the ON bits are building neatly. Probably GCC can do that. >> I will try with Sun Studio 12.1 >> with all latest patches. I have no idea why people use 12.1 when that >> compiler is unable to build simple things like GNU make or libgcrypt >> without blowing up on its own intermediate assembly[1] output but it >> seems >> able to build ON. My disdain for Studio 12.1 not withstanding it does >> seem >> to be the compiler to use. > > Not yet for building the OS. Studio 12 is still required - we're testing > a new patchset of Studio 12.1 now for building the OS and hope to be able > to switch to it soon, but until then, the official build environment is > still Studio 12 (except that ON uses lint, and only lint, from 12u1). > > http://hub.opensolaris.org/bin/view/Community+Group+tools/sun_studio_12_tools > has the set to use for now. Well I'll be using 12.1 entirely for ON and see what happens. Maybe it works. Maybe it fails. I would be far happier at this point with a decent GCC but my Solaris Application Dev books from Mr. Gove and the Solaris Internals books from McDougall and Mauro et. al. tell me the Studio compilers are the way to go. I do believe them even if it is somewhat mysterious, closed and proprietary. -- Dennis ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
Dennis Clarke wrote: > Getting to b144 may be a challenge as I do not know what patch levels are > required or even what compiler to use. As always, it's documented on http://hub.opensolaris.org/bin/view/downloads/on > I will try with Sun Studio 12.1 > with all latest patches. I have no idea why people use 12.1 when that > compiler is unable to build simple things like GNU make or libgcrypt > without blowing up on its own intermediate assembly[1] output but it seems > able to build ON. My disdain for Studio 12.1 not withstanding it does seem > to be the compiler to use. Not yet for building the OS. Studio 12 is still required - we're testing a new patchset of Studio 12.1 now for building the OS and hope to be able to switch to it soon, but until then, the official build environment is still Studio 12 (except that ON uses lint, and only lint, from 12u1). http://hub.opensolaris.org/bin/view/Community+Group+tools/sun_studio_12_tools has the set to use for now. -- -Alan Coopersmith-alan.coopersm...@oracle.com Oracle Solaris Platform Engineering: X Window System ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
> Alan Coopersmith wrote: >> > CPIO is limited to 8 GB files and the (non-POSIX ... SVR4) CPIO >> archive format >> > Solaris is currently using by default is even limited to 4 GB file >> size. >> >> So? If the OS installer is unpacking any files larger than 4gb, >> there is >> something seriously wrong. > > Are you sure about the size of the Xorg binary in 5 years? I don't think that a complete Xorg implementaion is that large at all. However, you and I were chatting about these things and I think for now the focus should be on the lower level issues like basic ON and utils and a compiler. >> > It has been shown already that pkgadd can be enhanced to use other >> > archive types and I am strongly against introducing something new >> > that is based n the outdated (since y-2001) CPIO. >> >> Using cpio in the installs is hardly new. > > I was replying to a person who proposed to use cpio for something new. no. Just .. no. > I know that some people at Sun did like to use cpio because they rely on > undocumented features in the Sun cpio implementation. I recommend to > better > use "star -install ..." as this is a documented feature that grants you to > replace life existing binaries wihout a crash. We both know that works very well and it is open and standards compliant. About your issues with lint and b134. I was able to build b138 without too much problem but I started incrementatlly from the last SXCE b130 and then went to b134 and then b138. You have access to that server in the Blastwave server farm and so you know this already. I install the compilers into a somewhat non-standard path and then make a symlink hack to adjust search paths. This has worked well for me. Getting to b144 may be a challenge as I do not know what patch levels are required or even what compiler to use. I will try with Sun Studio 12.1 with all latest patches. I have no idea why people use 12.1 when that compiler is unable to build simple things like GNU make or libgcrypt without blowing up on its own intermediate assembly[1] output but it seems able to build ON. My disdain for Studio 12.1 not withstanding it does seem to be the compiler to use. -- Dennis [1] I have seen this over and over and its just sad really. /opt/studio/SOS12.1/SUNWspro/bin/cc -DHAVE_CONFIG_H -I. \ -I../../../../libgcrypt-1.4.6/cipher -I.. -I../src \ -I../../../../libgcrypt-1.4.6/src -I/opt/csw/include \ -I/opt/csw/include -m64 -xarch=sse2 -erroff=%all -g -nofstore \ -Qy -Xa -xbuiltin=%none -xmodel=medium -xnolibmil -xnolibmopt \ -Kpic -xildoff -xregs=no%frameptr -xs -xstrconst \ -D_TS_ERRNO -c ../../../../libgcrypt-1.4.6/cipher/cipher.c \ -fPIC -DPIC -o .libs/cipher.o Assembler: "/tmp/yabeAAAmcaq8V", line 5753 : Syntax error Near line: "movl %eax" "/tmp/yabeAAAmcaq8V", line 5831 : Syntax error Near line: "movl %eax" "/tmp/yabeAAAmcaq8V", line 5915 : Syntax error Near line: "movl %eax" "/tmp/yabeAAAmcaq8V", line 6016 : Syntax error Near line: "movl %eax" "/tmp/yabeAAAmcaq8V", line 6114 : Syntax error Near line: "movl %eax" "/tmp/yabeAAAmcaq8V", line 6191 : Syntax error Near line: "movl %eax" "/tmp/yabeAAAmcaq8V", line 6275 : Syntax error Near line: "movl %eax" "/tmp/yabeAAAmcaq8V", line 6376 : Syntax error Near line: "movl %eax" "/tmp/yabeAAAmcaq8V", line 6561 : Syntax error Near line: "movl%edx, " Failure in /opt/sunstudio12.1/prod/bin/fbe, status = 0x7f00 Fatal Error exec'ing /opt/sunstudio12.1/prod/bin/fbe cc: acomp failed for ../../../../libgcrypt-1.4.6/cipher/cipher.c gmake[2]: *** [cipher.lo] Error 1 gmake[2]: Leaving directory `/export/medusa/dclarke/build/libgcrypt/i386/1.4.6/amd64_SOS12.1_try2/cipher' gmake[1]: *** [all-recursive] Error 1 gmake[1]: Leaving directory `/export/medusa/dclarke/build/libgcrypt/i386/1.4.6/amd64_SOS12.1_try2' gmake: *** [all] Error 2 This is nothing new to me. Studio 12 works but 12.1 fails in so many ways in so many places it is just scary. GCC 4.x works fine. ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
joerg.schill...@fokus.fraunhofer.de wrote: > Alan Coopersmith wrote: > >> joerg.schill...@fokus.fraunhofer.de wrote: >>> Alan Coopersmith wrote: >>> > CPIO is limited to 8 GB files and the (non-POSIX ... SVR4) CPIO archive > format > Solaris is currently using by default is even limited to 4 GB file size. So? If the OS installer is unpacking any files larger than 4gb, there is something seriously wrong. >>> Are you sure about the size of the Xorg binary in 5 years? >> -r-sr-xr-x 1 root bin 2.9M Jul 13 10:01 /usr/bin/amd64/Xorg* >> >> I can't imagine it growing 2048 times larger in the near future. > > We will never need more than 640 G of RAM. I didn't say "never" - but not in the "Solaris Next timescale". >> Having files that big would certainly kill the CD media installer, and >> probably >> the DVD one too, but we're not yet ready to move to BluRay install media. > > Solaris can of course use BluRay media as there is cdrtools. But so few machines on which it will be installed have BluRay drives that it's not yet an option for the install media itself. Fortunately, we don't need to worry about that yet. -- -Alan Coopersmith-alan.coopersm...@oracle.com Oracle Solaris Platform Engineering: X Window System ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
Alan Coopersmith wrote: > joerg.schill...@fokus.fraunhofer.de wrote: > > Alan Coopersmith wrote: > > > >>> CPIO is limited to 8 GB files and the (non-POSIX ... SVR4) CPIO archive > >>> format > >>> Solaris is currently using by default is even limited to 4 GB file size. > >> So? If the OS installer is unpacking any files larger than 4gb, there is > >> something seriously wrong. > > > > Are you sure about the size of the Xorg binary in 5 years? > > -r-sr-xr-x 1 root bin 2.9M Jul 13 10:01 /usr/bin/amd64/Xorg* > > I can't imagine it growing 2048 times larger in the near future. We will never need more than 640 G of RAM. > Having files that big would certainly kill the CD media installer, and > probably > the DVD one too, but we're not yet ready to move to BluRay install media. Solaris can of course use BluRay media as there is cdrtools. There is a smaller problem with the fact that there is no large file support in ISO-9660, but there is still large file support in the UDF implementation of Solaris. With cdrtools, you may already create a 50 GB install medium and if the BDX spec is compatible to BD, this will even work with media up to 128 GB. Jörg -- EMail:jo...@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin j...@cs.tu-berlin.de(uni) joerg.schill...@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
joerg.schill...@fokus.fraunhofer.de wrote: > Alan Coopersmith wrote: > >>> CPIO is limited to 8 GB files and the (non-POSIX ... SVR4) CPIO archive >>> format >>> Solaris is currently using by default is even limited to 4 GB file size. >> So? If the OS installer is unpacking any files larger than 4gb, there is >> something seriously wrong. > > Are you sure about the size of the Xorg binary in 5 years? -r-sr-xr-x 1 root bin 2.9M Jul 13 10:01 /usr/bin/amd64/Xorg* I can't imagine it growing 2048 times larger in the near future. Having files that big would certainly kill the CD media installer, and probably the DVD one too, but we're not yet ready to move to BluRay install media. >>> It has been shown already that pkgadd can be enhanced to use other >>> archive types and I am strongly against introducing something new >>> that is based n the outdated (since y-2001) CPIO. >> Using cpio in the installs is hardly new. > > I was replying to a person who proposed to use cpio for something new. No, he was simply referring to the OS install cd's, which have used cpio for a long time. -- -Alan Coopersmith-alan.coopersm...@oracle.com Oracle Solaris Platform Engineering: X Window System ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
Alan Coopersmith wrote: > > CPIO is limited to 8 GB files and the (non-POSIX ... SVR4) CPIO archive > > format > > Solaris is currently using by default is even limited to 4 GB file size. > > So? If the OS installer is unpacking any files larger than 4gb, there is > something seriously wrong. Are you sure about the size of the Xorg binary in 5 years? > > It has been shown already that pkgadd can be enhanced to use other > > archive types and I am strongly against introducing something new > > that is based n the outdated (since y-2001) CPIO. > > Using cpio in the installs is hardly new. I was replying to a person who proposed to use cpio for something new. I know that some people at Sun did like to use cpio because they rely on undocumented features in the Sun cpio implementation. I recommend to better use "star -install ..." as this is a documented feature that grants you to replace life existing binaries wihout a crash. Jörg -- EMail:jo...@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin j...@cs.tu-berlin.de(uni) joerg.schill...@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
joerg.schill...@fokus.fraunhofer.de wrote: > andrew wrote: >> So instead of OpenSolaris installing just one CD's worth via CPIO, you will >> have a Solaris Next DVD running the Caiman installer doing CPIO-based >> installations of a whole DVD worth of IPS packages. Or perhaps Oracle will >> manage to get an IPS-based installer working fast enough to be viable as the >> OS installer itself, which would enable more fine-grained control of which >> software to install from the DVD. > Why are you interested in CPIO? > > CPIO is limited to 8 GB files and the (non-POSIX ... SVR4) CPIO archive format > Solaris is currently using by default is even limited to 4 GB file size. So? If the OS installer is unpacking any files larger than 4gb, there is something seriously wrong. > It has been shown already that pkgadd can be enhanced to use other > archive types and I am strongly against introducing something new > that is based n the outdated (since y-2001) CPIO. Using cpio in the installs is hardly new. -- -Alan Coopersmith-alan.coopersm...@oracle.com Oracle Solaris Platform Engineering: X Window System ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
Re: [osol-discuss] Solaris Next timescale
andrew wrote: > So the big question is: will Solaris Next use IPS packaging? My money is on > IPS packaging but a big fat CPIO-based DVD installation similar to S10 but > using the new Caiman installers from OpenSolaris (local text and > graphics-based installs plus via serial line, network and service processor > consoles). > > So instead of OpenSolaris installing just one CD's worth via CPIO, you will > have a Solaris Next DVD running the Caiman installer doing CPIO-based > installations of a whole DVD worth of IPS packages. Or perhaps Oracle will > manage to get an IPS-based installer working fast enough to be viable as the > OS installer itself, which would enable more fine-grained control of which > software to install from the DVD. > > As I said above, this is pure speculation based on some of the noises coming > out of Oracle right now. In my view Sun waited too long between the release > of Solaris 10 and Solaris Next. Solaris Next should have been released in > 2009, in my view. Why are you interested in CPIO? CPIO is limited to 8 GB files and the (non-POSIX ... SVR4) CPIO archive format Solaris is currently using by default is even limited to 4 GB file size. It has been shown already that pkgadd can be enhanced to use other archive types and I am strongly against introducing something new that is based n the outdated (since y-2001) CPIO. Jörg -- EMail:jo...@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin j...@cs.tu-berlin.de(uni) joerg.schill...@fokus.fraunhofer.de (work) Blog: http://schily.blogspot.com/ URL: http://cdrecord.berlios.de/private/ ftp://ftp.berlios.de/pub/schily ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
[osol-discuss] Solaris Next timescale
Time for some crystal ball gazing folks. I reckon beta 1 of Solaris Next will be announced at Oracle Open World. Beta 2 will follow within 3 to 6 months, and by this time next year we will have Solaris Next FCS. I'm hoping Oracle are currently working on providing a smooth migration path from Solaris 10 to Solaris Next that doesn't involve dumping an existing S10 system into an S10 branded zone on Sol_Next. So the big question is: will Solaris Next use IPS packaging? My money is on IPS packaging but a big fat CPIO-based DVD installation similar to S10 but using the new Caiman installers from OpenSolaris (local text and graphics-based installs plus via serial line, network and service processor consoles). So instead of OpenSolaris installing just one CD's worth via CPIO, you will have a Solaris Next DVD running the Caiman installer doing CPIO-based installations of a whole DVD worth of IPS packages. Or perhaps Oracle will manage to get an IPS-based installer working fast enough to be viable as the OS installer itself, which would enable more fine-grained control of which software to install from the DVD. As I said above, this is pure speculation based on some of the noises coming out of Oracle right now. In my view Sun waited too long between the release of Solaris 10 and Solaris Next. Solaris Next should have been released in 2009, in my view. Cheers Andrew. -- This message posted from opensolaris.org ___ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
