From tpchan at comcast.net Thu Dec 3 22:23:13 2009 From: tpchan at comcast.net (Terry Chan) Date: Thu, 3 Dec 2009 15:23:13 -0600 Subject: Git moonbase needs a fresh clone Message-ID: <20091203212313.GA27968@sega.hsd1.il.comcast.net> To all lunar devs with git push rights on doppio. Due to a recent problem on doppio, you MUST get a NEW copy of the git moonbase with: git clone git://lunar-linux.org/lunar/moonbase.git I had to fix the problem on doppio with this command: git update-refs HEAD 1ed9c2debf5361b9b26b03d611a4cdb45cb5f0ed If you fail to clone a new moonbase your current git moonbase will be in an incorrect state if you have ANY of the elangelo changes in your repo from the last few days. Things to search for are "lvu DETAILS xfce4-panel" and the version number should be 4.6.1, not git- something or other. Also you should NOT have a copy of zbeta/mplayer in your repo. If you have either of those two cases then your repo is messed up and git pull will not correctly fix your repo. You MUST do a fresh git clone and get a brand spankin' new moonbase git repo for yourself. Terry Chan From tpchan at comcast.net Thu Dec 3 22:27:41 2009 From: tpchan at comcast.net (Terry Chan) Date: Thu, 3 Dec 2009 15:27:41 -0600 Subject: Git moonbase mail Message-ID: <20091203212741.GA28111@sega.hsd1.il.comcast.net> It was just pointed out to me that if you don't get a new clone moonbase with: git clone git://lunar-linux.org/lunar/moonbase.git Then you do have a choice on your local repo to do this command: git reset --hard 1ed9c2debf5361b9b26b03d611a4cdb45cb5f0ed That should get you to the state before the elangelo merge and your local repos should then be able to do git pull to update correctly again. Still check for the xfce4-panel/DETAILS for version 4.6.1. Also check for zbeta/mplayer <--- should not exist in your repo. Terry Chan From ratler at lunar-linux.org Fri Dec 4 09:06:38 2009 From: ratler at lunar-linux.org (Stefan Wold) Date: Fri, 04 Dec 2009 09:06:38 +0100 Subject: Git moonbase needs a fresh clone In-Reply-To: <20091203212313.GA27968@sega.hsd1.il.comcast.net> References: <20091203212313.GA27968@sega.hsd1.il.comcast.net> Message-ID: <1259913998.4056.4.camel@localhost> Hey fellow developers, I have now revoked everyones commit access until you can verify that you have a clean moonbase, first then I will re-add your commit access. We have managed to push out the bad changes a couple of times more since we fixed the issue. I'm sorry for any inconvenience but this may cause but right now it seem to be the best solution. -- Sincerely Stefan Wold Lunar Linux developer - PGP public key 6E810F05 On Thu, 2009-12-03 at 15:23 -0600, Terry Chan wrote: > To all lunar devs with git push rights on doppio. > > Due to a recent problem on doppio, you MUST get a NEW copy of the git moonbase > with: > > git clone git://lunar-linux.org/lunar/moonbase.git > > I had to fix the problem on doppio with this command: > > git update-refs HEAD 1ed9c2debf5361b9b26b03d611a4cdb45cb5f0ed > > > If you fail to clone a new moonbase your current git moonbase will be in an > incorrect state if you have ANY of the elangelo changes in your repo from the > last few days. > > Things to search for are "lvu DETAILS xfce4-panel" and the version number should > be 4.6.1, not git- something or other. > > Also you should NOT have a copy of zbeta/mplayer in your repo. > > If you have either of those two cases then your repo is messed up and git pull > will not correctly fix your repo. You MUST do a fresh git clone and get a brand > spankin' new moonbase git repo for yourself. > > Terry Chan > _______________________________________________ > Lunar-dev mailing list > Lunar-dev at lunar-linux.org > http://foo-projects.org/mailman/listinfo/lunar-dev -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 198 bytes Desc: This is a digitally signed message part URL: From zbiggy at o2.pl Sat Dec 19 03:15:08 2009 From: zbiggy at o2.pl (Zbigniew Luszpinski) Date: Sat, 19 Dec 2009 03:15:08 +0100 Subject: gcc 4.4 plugin needs further update for Alpha, PowerPC, Sparc and verify Intel extensions Message-ID: <200912190315.08556.zbiggy@o2.pl> Hello, I have just updated gcc 4.4 plugin. See changes and how to install on General Lunar mailing list. The things which need some more work are: *CPU families I do not know: Alpha, PowerPC, Sparc - someone who knows have to update CPU list for these architectures. x86 and x86_64 families are current now. *Intel CPU extensions like: SSSE3 SSE4 SSE4.1 SSE4.2 SSE5 - this is for future AMD CPUs. No present CPU has this. AVX AES PCLMUL I do not have Intel CPU so these extensions are not detected now. Someone who has recent Intel CPU can check /proc/cpuinfo and update gcc plugin or send me /proc/cpuinfo file so I will look and update gcc plugin by myself. have a nice day, Zbigniew Luszpinski From tpchan at comcast.net Sat Dec 19 16:11:06 2009 From: tpchan at comcast.net (Terry Chan) Date: Sat, 19 Dec 2009 09:11:06 -0600 Subject: Your changes to gcc_4_4 plugin are incorrect Message-ID: <20091219151106.GA14136@sega.hsd1.il.comcast.net> zbiggy, You really need to test important changes like this MUCH more before committing to moonbase. gcc-4.4.2 installed, glibc-2.10.1, kernel-2.6 Downloaded latest plugin, installed into /var/lib/lunar/plugins, lunar optimize and set to "autodetect". Contents of my /etc/lunar/local/optimizations.GCC_4_4 file: CPU=autodetect BOPT=Faster SPD=( Pointers ) XTRA=( MMX SSE SSE2 dnow ) FPM=Both CC_OPTS=( Pipe ) STACK= Attached is the compile log for the 1st module I tried to compile with your new plugin. Terry Chan ------------ -------------- next part -------------- A non-text attachment was scrubbed... Name: git-1.6.5.7.bz2 Type: application/octet-stream Size: 469 bytes Desc: not available URL: From tpchan at comcast.net Sat Dec 19 16:19:04 2009 From: tpchan at comcast.net (Terry Chan) Date: Sat, 19 Dec 2009 09:19:04 -0600 Subject: autodetect in gcc-4.4.2 Message-ID: <20091219151904.GA30961@sega.hsd1.il.comcast.net> As I read the info pages for gcc-4.4.2, I find absolutely NO reference to autodetect as a valid parameter to -march nor -mtune. Please fix the changes you made to the optimize plugin. Terry Chan From zbiggy at o2.pl Sat Dec 19 19:43:45 2009 From: zbiggy at o2.pl (Zbigniew Luszpinski) Date: Sat, 19 Dec 2009 19:43:45 +0100 Subject: Your changes to gcc_4_4 plugin are incorrect In-Reply-To: <20091219151106.GA14136@sega.hsd1.il.comcast.net> References: <20091219151106.GA14136@sega.hsd1.il.comcast.net> Message-ID: <200912191943.46177.zbiggy@o2.pl> > zbiggy, > > You really need to test important changes like this MUCH more before committing > to moonbase. I commited wrong file :( I have just commited the right one. Please try it and let me know if it works for you. The only difference is 2 words were misplaced in line. P.S. To make it active please select any other CPU and again native/autodetect to fix Lunar's etc settings. have a nice day, Zbigniew Luszpinski From zbiggy at o2.pl Sun Dec 20 01:04:18 2009 From: zbiggy at o2.pl (Zbigniew Luszpinski) Date: Sun, 20 Dec 2009 01:04:18 +0100 Subject: Yet another new update to gcc plugin - devel feature added Message-ID: <200912200104.18379.zbiggy@o2.pl> Hello, in second update to gcc plugin I made some tiny tweaks, added Intel extensions (see post on LGML) and added one feature designed only for more comfortable development: $MYCFLAGS You can define $MYCFLAGS to overwrite Lunar menu settings. This little feature will help us fast testing different gcc flags without the need of changing module or browsing lunar menu. You know how to make right use of it and what benefits it provides. I do not provide wider description because I would like to see this as development tool only - not everyday use feature for everybody. Use it only for testing. Unset before doing final build to restore Lunar settings. have a nice day, Zbigniew Luszpinski From jean.bruenn at ip-minds.de Sun Dec 20 20:44:11 2009 From: jean.bruenn at ip-minds.de (Jean-Michel Bruenn) Date: Sun, 20 Dec 2009 20:44:11 +0100 Subject: Issues with mfpmath=both Message-ID: <4B2E7E8B.3000706@ip-minds.de> Hi, i'm currently running gcc 4.2, today i wanted to update my whole box and ran into problems due to mfpmath=both. That switch is not there (you can see it in the output later in this email). So people upgrading will have trouble compiling "anything" because it's simply not working with that switch, IF they use mfpmath=both. For GCC 4.4 it's the same - There's no option "both" - I'm wondering who brought that in, however, i guess this should be fixed as soon as possible. configure:4060: checking for C compiler default output file name configure:4082: gcc -O2 -march=athlon64 -mmmx -msse -msse2 -msse3 -m3dnow -mfpmath=both -pipe -s conftest.c >&5 conftest.c:1: error: bad value (both) for -mfpmath= switch configure:4086: $? = 1 configure:4123: result: configure: failed program was: | /* confdefs.h */ | #define PACKAGE_NAME "" | #define PACKAGE_TARNAME "" ye somehow someone made that lunar is using "both" if you have selected "both" in lunar optimize and "both" is not working, thus compiles arent working. Dunno whether this issue is just dedicated to me. :-) Cheers Jean From jean.bruenn at ip-minds.de Sun Dec 20 21:05:48 2009 From: jean.bruenn at ip-minds.de (Jean-Michel Bruenn) Date: Sun, 20 Dec 2009 21:05:48 +0100 Subject: Issues with mfpmath=both In-Reply-To: <4B2E7E8B.3000706@ip-minds.de> References: <4B2E7E8B.3000706@ip-minds.de> Message-ID: <4B2E839C.4070006@ip-minds.de> Hi, correction: 4.4 lists it' just 4.2 not. So ppl upgrading from gcc below 4.4 will have trouble to upgrade. And ppl below gcc 4.4 will have trouble to compile anything IF they use "both". wdp__, as far as concerne the mfpmath=both issue i have to say that gcc-4.4.2 man page list the 'both' choice for the mfpmath option but also informs that it is an experimental feature Cheers Jean-Michel Bruenn wrote: > Hi, > > i'm currently running gcc 4.2, today i wanted to update my whole box > and ran into problems due to mfpmath=both. That switch is not there > (you can see it in the output later in this email). So people > upgrading will have trouble compiling "anything" because it's simply > not working with that switch, IF they use mfpmath=both. For GCC 4.4 > it's the same - There's no option "both" - I'm wondering who brought > that in, however, i guess this should be fixed as soon as possible. > > configure:4060: checking for C compiler default output file name > configure:4082: gcc -O2 -march=athlon64 -mmmx -msse -msse2 -msse3 > -m3dnow -mfpmath=both -pipe -s conftest.c >&5 > conftest.c:1: error: bad value (both) for -mfpmath= switch > configure:4086: $? = 1 > configure:4123: result: > configure: failed program was: > | /* confdefs.h */ > | #define PACKAGE_NAME "" > | #define PACKAGE_TARNAME "" > > ye > somehow someone made that lunar is using "both" > if you have selected "both" in lunar optimize > and "both" is not working, thus compiles arent working. > > Dunno whether this issue is just dedicated to me. :-) > > Cheers > Jean > _______________________________________________ > Lunar-dev mailing list > Lunar-dev at lunar-linux.org > http://foo-projects.org/mailman/listinfo/lunar-dev From zbiggy at o2.pl Mon Dec 21 00:53:49 2009 From: zbiggy at o2.pl (Zbigniew Luszpinski) Date: Mon, 21 Dec 2009 00:53:49 +0100 Subject: Issues with mfpmath=both In-Reply-To: <4B2E839C.4070006@ip-minds.de> References: <4B2E7E8B.3000706@ip-minds.de> <4B2E839C.4070006@ip-minds.de> Message-ID: <200912210053.49510.zbiggy@o2.pl> > Hi, > > correction: 4.4 lists it' just 4.2 not. So ppl upgrading from gcc below > 4.4 will have trouble to upgrade. And ppl below gcc 4.4 will have > trouble to compile anything IF they use "both". > > wdp__, as far as concerne the mfpmath=both issue i have to say > that gcc-4.4.2 man page list the 'both' choice for the mfpmath option > but also informs that it is an experimental feature > I introduced 'both' parameter because some modules breaks because of comma between 387,sse. But only for 4.4 plugin. The 4.2 and lower plugins are untouched because both parameter is not present there. Do you have gcc 4.2 plugin set as default? Did you try to re set fpu parameter after switching default compiler to gcc 4.2? It is possible you have found bug in Lunar's plugin management. It would be very bad thing if lunar update would set default gcc plugin to 4.4 if gcc 4.2 is still in use. If you confirm this bug I will look at it more closely and provide some kind of fix to keep gcc and plugin version bound together. You can now have other problems with features which came with gcc4.4 like phenom, autodetect, new Intel sse... not present in older gcc. Make sure in Lunar menu you have set as default the right plugin. Gcc plugin and gcc version must equal to match features they provide. Experimental feature is to use sse and 387 together. Such warning is with every gcc release, I remember it since gcc3. The 'both' word is just rename of sse,387 to eliminate comma problem in some build scripts. I always used sse,387 and now both parameter. So far all modules work perfect. have a nice day, Zbigniew Luszpinski From jean.bruenn at ip-minds.de Mon Dec 21 02:09:19 2009 From: jean.bruenn at ip-minds.de (Jean-Michel Bruenn) Date: Mon, 21 Dec 2009 02:09:19 +0100 Subject: Issues with mfpmath=both In-Reply-To: <200912210053.49510.zbiggy@o2.pl> References: <4B2E7E8B.3000706@ip-minds.de> <4B2E839C.4070006@ip-minds.de> <200912210053.49510.zbiggy@o2.pl> Message-ID: <4B2ECABF.9040600@ip-minds.de> > I introduced 'both' parameter because some modules breaks because of comma between 387,sse. > But only for 4.4 plugin. The 4.2 and lower plugins are untouched because both parameter is not present there. > Do you have gcc 4.2 plugin set as default? Did you try to re set fpu parameter after switching default compiler to gcc 4.2? > It is possible you have found bug in Lunar's plugin management. It would be very bad thing if lunar update would set default gcc plugin to 4.4 if gcc 4.2 is still in use. If you confirm this bug I will look at it more closely and provide some kind of fix to keep gcc and plugin version bound together. You can now have other problems with features which came with gcc4.4 like phenom, autodetect, new Intel sse... not present in older gcc. Make sure in Lunar menu you have set as default the right plugin. Gcc plugin and gcc version must equal to match features they provide. > as i fixed my system manually now, i can't test it anymore. However, the manpage even states you can use "sse+387". I never had any problems with 387,sse and i'm using those settings since i started ago some years with lunar. So let's hope nobody else comes into such trouble.