From: UnixOS2 Archive To: "UnixOS2 Archive" Date: Sun, 29 Sep 2002 04:37:18 EST-10EDT,10,-1,0,7200,3,-1,0,7200,3600 Subject: [UnixOS2_Archive] No. 331 ************************************************** Saturday 28 September 2002 Number 331 ************************************************** Subjects for today 1 Re: Summary: gcc 3.0.3 testing : John Poltorak 2 Re: Summary: gcc 3.0.3 testing : Steve Wendt" 3 Re: Summary: gcc 3.0.3 testing : DWParsons at t-online.de (Dave Parsons) 4 Re: Summary: gcc 3.0.3 testing : Thomas Dickey 5 Re: Summary: gcc 3.0.3 testing : Andrew Belov" 6 Re: Summary: gcc 3.0.3 testing : Steve Wendt" 7 Re: Summary: gcc 3.0.3 testing : Andrew Belov" 8 Re: Summary: gcc 3.0.3 testing : Franz Bakan" 9 Re: Summary: gcc 3.0.3 testing : DWParsons at t-online.de (Dave Parsons) **= Email 1 ==========================** Date: Sun, 29 Sep 2002 10:29:54 +0100 From: John Poltorak Subject: Re: Summary: gcc 3.0.3 testing On Sat, Sep 28, 2002 at 09:04:27PM +0100, Thomas Hoffmann wrote: > Since it is rather quiet here, I thought I send a summary of my testing > gcc 3.0.3. I put an emx0.9df4/ gcc 3.0.3 combo into a FHS like directory > structure and ran the tests from the emx test directory. > > Maybe somebody can comment the results .... I wanted to standardise on gcc 3.0.3 but never managed to build Perl with it so gave up. I'd love to make to some progress on that front but have never been able to sort out the problems. Has anyone else managed to build Perl using gcc 3.0.3 ? -- John **= Email 2 ==========================** Date: Sun, 29 Sep 2002 14:58:22 -0700 (PDT) From: "Steve Wendt" Subject: Re: Summary: gcc 3.0.3 testing On Sun, 29 Sep 2002 18:42:26 +0300 (MSK), Andrew Belov wrote: >Still waiting for a decent GCC v 3.1.x port, though. :-( Why not 3.2? 3.1.x is dead? ----------- "Good people do not need laws to tell them to act responsibly, while bad people will find a way around the laws." - Plato (427-347 B.C.) **= Email 3 ==========================** Date: Sun, 29 Sep 2002 18:01:41 +0200 (CES) From: DWParsons at t-online.de (Dave Parsons) Subject: Re: Summary: gcc 3.0.3 testing On Sun, 29 Sep 2002 18:42:26 +0300 (MSK), Andrew Belov wrote: < big snip > > Still waiting for a decent GCC v 3.1.x port, though. :-( > Have you managed to get any kind of complete 3.1 port? I got through stage 1 a few months back but I couldn't get it to bootstrap. -- Dave **= Email 4 ==========================** Date: Sun, 29 Sep 2002 18:24:30 -0400 From: Thomas Dickey Subject: Re: Summary: gcc 3.0.3 testing On Sun, Sep 29, 2002 at 02:58:22PM -0700, Steve Wendt wrote: > On Sun, 29 Sep 2002 18:42:26 +0300 (MSK), Andrew Belov wrote: > > >Still waiting for a decent GCC v 3.1.x port, though. :-( > > Why not 3.2? 3.1.x is dead? when someone bumps the version number suddenly like that, I tend to assume that the newer version is the one that's unstable. -- Thomas E. Dickey http://invisible-island.net ftp://invisible-island.net **= Email 5 ==========================** Date: Sun, 29 Sep 2002 18:42:26 +0300 (MSK) From: "Andrew Belov" Subject: Re: Summary: gcc 3.0.3 testing --_=_=_=IMA.BOUNDARY.H381QP138764=_=_=_ Content-Type: text/plain; charset="koi8-r" Content-Transfer-Encoding: 7bit On Sun, 29 Sep 2002 10:29:54 +0100, John Poltorak wrote: >I wanted to standardise on gcc 3.0.3 but never managed to build Perl with >it so gave up. I'd love to make to some progress on that front but have >never been able to sort out the problems. > >Has anyone else managed to build Perl using gcc 3.0.3 ? Not experimented with PERL, but with some other software (Lynx, etc.) there are two general issues: 1. To work around empty a.out files not recognized by binutils, use the enclosed forwarder (rename the original GCC.EXE to GCC2.EXE, then copy GCC.EXE from archive to its former location). It inserts one empty static variable into empty files, so they can be linked normally. 2. "-O2" is unsafe, whereas "-O2 -fno-expensive-optimizations" handles quite complex code correctly. "-O2 -fno-expensive-optimizations -funroll-loops -fpeephole2 -march=athlon" works for me in several projects. Still waiting for a decent GCC v 3.1.x port, though. :-( --_=_=_=IMA.BOUNDARY.H381QP138764=_=_=_ Content-Type: application/octet-stream; name="gcc_fwd.zip" Content-Transfer-Encoding: base64 UEsDBBQAAgAIAA6XNi3MDkJRyAYAAMoHAAAHAFYAZ2NjLmV4ZQkAQQA8AAAACACViTeUs2FgYJAB YgYODgY9x4CAkMgAV4Z//1kYmBiYHPRCIkFSDKx8DHpgib//uRhcK1KTS0sSk3JSAVVUDQAHLOiN PZ4dlz0i6I09bVQNUBTnGf6+3f3uvuMOb0+O5MCj7CF/ViAcjCM4xHLAiloORcDQgEGUM6iRw/NO 0ErqstzAZXst18pE8jOTa8K0omEggxO5SaaHhxCT2IQm0XRoZ0y0E26WNkRsIC2yXQRsMtN35pn3 Z5/3ed/5duYtqQIPTfwFXA4IAEjZLWUMWo7Bfhn0yucV2tllB0ZlUDIS5fo2AjyyPDkvhP/Li+S4 QsZm4uGIR9xV7w0/poxe4VLgh/ZIdj/4v/adEjzcGDHLe6+O1ax0k8yyBvyBDvmwTn5P542V6av1 VX6TjFsUeqS5hHqAAHmouX619873gGE1QeRDDBspSMjuFAURTOUIuI0k7yqUXBeCY5j7GMDtGIYp qJIpr1DwfYi51xFUY64OwQ8xjKOgFyphDYJWqOQSEfzLksTfOqj7lBKOIHhQbgNKaIOpkCBS4c/R vxDmPoLwMOaSSXgEcxcR/DHmigE8msoh+ByEKfAoDMPXCMxtRjADc98R0L6kOfMkdYnCXB6CZszd RDBrqegchkdlpgXBIcyFEbyM4U8IIkdeUqOErahkR0FhOS7aVV5oKSkpBwoI5pQP3+sslv0a0DEL NCoCUKxCgYiks7W2Y03OU7V1dpezttnuOFrnsLsa62sPP9tod9hqj9lCs2QmePbgwawMW4sNZNhN OC39IEi3A8cB0Hygbm1mOVtRuduypzjDUm4FwFgpTPPTmnJIZHpY7KnUeKz0ODsv/7tXyXF2YQPJ haLJkTw+CARWk8sutKYGmgDFKRhasC4IVOeUTjtJkuqgUyfG0+KPvGd9iBYmvR4TlcYu5HvYBfFt YRKZsS6VPmO4AjmFigw38iRpKdR784H35eOGSIXW+6mHGCauFwVOuIv29RadT+OD0rpgvjqobW80 xYzsMbS1EMAZP6Jk9J38VUJgsd+gJhVibnfT9msbHj8WpT9fgKKizBXdO8ZMFb8lI7KD4WELq2/L SXcavkgoS4D9PsbCv0hqwxmSJCXyrfNAey6YJA3+YVMObcvZeyE+kGFOSfOw9/nW+8AVGzjsb7NU JL1pMG+3RCZF3v+EjM0OilqGH9VUC3/fF2pjmLH2xHceSBLQUYj6B81N716ryzppelU15v4k1ev+ /J+StNvr/quP3LBxgIhTet23PZUNVC8t6KUmPECa1oTpQM3gDdWZUb2Bb10AruhB9U3m8k7dPq+3 iLlQHdjfG7uQKFg1QiX2uj/lLJ6yqZ8OziTt7LSsZ9x3v5akcXbG96K5u+w9I85lZ7Reva9Uz2oE 68zvGGsxP01Ze6mJ24nzAjsjWOc9vzQl4QDzJ8Zs4A8k5c997mGn+NYp4HxMYOc9bt85WLUBzQvr qI1jIGGcnTYnXKyHgnW6LQ8DbfvTuSkJLltKUIfS1WOgplI9GtpD0i9osODX6DuxwE5Dbfp/psXu 0FWVLeU9HYKl8DjEhpnw1KIkESm/SjGweLzdPcWd8ld3P++NIYVudUFydvDCYtu7cWFt+/voxknL mIn2IPeHMST/rga46LeScfjKmnRT5s4t6VHZfxbMX32DPvoSeNi7Rr71LqN1H4pd/8UT6jlF4JR7 s05NxoslMZHS5Z644bIxBm7DFPjSSg4peyr5USzSWWYlSWkI87eZPiQfnDZzNNXwJlkvbeohPUl+ 0zeEkZ+XXGuuYB2dFPo1vh7KSvbxFto4TQmsfqqLQ8aGIIKgAcK3yyLhkPs3dmMDlk+4EDkYQWuE 6POwxkP4z5XpR6o7eEVqZaafyTfwtwlXnF9nDE1sVaVNPW7kzyRi4EJ+lX8rdaQ/l3PqTsCmqvK9 Twnz2cFvlZJLx39NPLjXv1DcMWmGmq6nhu/lZUr3+ucmdggTZcLcH319rDYb0vhShGMviujYnsc4 4/NUznV5Nc6NeQnO8rmQLtrgUhbzH0CoNxSIn3XNhW4yaU6qQLzexd9ZFENQMxLVHUX5N6V2LM72 75qb6Fi8BNc3hwhETpSKr0yET9AFYrc52ejC/AdEuIVbp+fvPAg3QSrmJc+T4ltbyVvFI18R5C3h qp6fBy7m5tM9BAioHaaBes5NH8FHoEkR/drr+dmTP7MIHz9D/tuzhb9Kzb5R1lzWsqnl96hlPWqJ R5Wbr5Xae2PU5tPHTveJE9b9feI11fPOd/QDoE8cGon5LAYXUnTVQI1TOdtXKr7QMTkRnu07ETuQ 0Bc+7VurHCXlq6NjO6sQtduyxZ1D9ei6xhlQ+0z19VBXCABdcWEhc8juaK5z1NscDFvFygd1SBW4 iG+YbjC7yp/IYk4y2RmZjN3BPFfnlCkO23HXYYetPjIi8b9QSwMECgAAAAAAWpQ9LQAAAAAAAAAA AAAAAAQAEQBzcmMvVVQNAAecHZc9lh2XPZYdlz1QSwMEFAACAAgAVGozLc7Lsl8QBQAAcQsAAAkA EQBzcmMvZndkLmNVVA0AB4CkiT2mHZc9COiNPX1We0/bSBD/H4nvMOQq8nKMN3DlAnUloGmLClzv aHWVIIocZ0MWHK+7XpNyJ777zeyuHQcQhDjeec9vHvZOB96cTg9gtpz6sXcPzDcf6AdBfycY7LAB sOCA/XGwx0BJqWH4K4M30NnZ3Njc+E2kcVJMObzLlIx5nvvz93VqrqdCPqclYvKMqER6Y4hInvKZ SDmcXpycjY+PLod1SZn3ndhOBy6iBQc5A6nEjUijBD6dnPjDH0NYyGmRcB/Oi1zDhMMdzzSIFPSc Q05KU6F4rKV6gCgn6uYGAMykWkZqypXv8st1pEUM8TxScBPH4xRVr0ZhA+/7Pv/FG4frMnJyO+a/ NIn4kpgmystikmuhCy1k6iKDHtgkc4hS4ItMP4AzdB8pEU1QxIZQpLm4SfnUOsiLydhaQB//YdD7 HvOCVz6s/yo78PbcL9p6+4S1u+9uBr97LGD4HXgMDTL2Fu0yIg/QP0MO2zdUEmQDsmWpe6izb4X2 aoIYVuCs0nfXkioNZlkDc4u2gs2NxxLLI60JLayZhJsCO86UVBY6KzTMRMKpRB6IGeL64MPfXBcq zSEgyp9ffFPm4++fLg+w0YHqCFmk5/5S4GUqdX5d2ugRCe32iNqApUybGq/qzpiwf31ngnSYb3/7 flyXBo5muCIlW0+Rahv42AbdMnXtTHmuMWxkRuom9my1Ox083Lc3N6jSxBOH7iYe52FAByuXGXgA 8fkoFHZ8POfxHaW85GiPQywXmUhwwgg2bDjsU+x+E68Jizq/JdAiiHfkf3sbBzJeZC3yfyVGHjR6 caON7G63TbmYOGYt8T4kcUNSBuoWax/aSC7kEnAWVISOqEYlsC86rCxTpiYPWaTTLStLnrYWfPEk INnwoN82vo0WiTn2VX+0FTavg2bbVouyyR5aFmQn0+2bSAF4knOTjImkx17TYTUdK7aWdnUM7PGR LqX9MraAYus1a4G/5MrZo0KzytajhX0Lqc8xR0YrC9GWiufKGWv6zXY7DC++n50ZBfIUacd028oo lwm5ULJ1bj2rRzeJ5xHuU4VNTKt61dsLJLde62NLUZ7tWtPNKi7vhAe3HvwspOZTolXbL5HYvEmt 4yMVLfIrFvT3RhUVN7Jd0Scnn8+Pfnw9+vbZML9+Oz2GTIuJOeB9J8vsodraT9Vs/pniOU+1HbWP p2dD6CBAPFrYefsg809cn6YzeZzI+C5vbZMTD7bJvAEOJf4quHo4NzubHlct4vXe42U8X2hcVSun XpVAVU9V1bNkYU2vr9eLqsKSWa+k6nbp6AqKeJcPMGO8nt3W2j4q0fBcBU3pSMVBHozW3NlBZi8P cha69dZeRAlC1MJocAuUo9DuuBl0psUoND3hGiAkccq9GvomNNtbNvFDtxisqJ3ZTtbths1G0/Ao sFvaMOXc3bqdcAi3bos9Wxu3o5D03Qaw5hDtw5X1leRqwF+J4nGFG6HTZSMTvZmpOMyzaJneZ62v 43+OTr95K4TL27IPatWqsKX2sM0YzmSGoK4K11CTxlqHlAOsYovbLOf8zml7EHhwORx+GQ8vPlh2 Es40TxIn4FTiROZ8jWSWQII4L3BAIx/7p3yw2FeUHObRPaflL/DNRfzLIQxht+85SixTzWnm8wP3 TA32IWAQBO7f9336qZ4BSRju9muFK1J8pN1VebuN+Rouy6e4PAUGX4OV0Jhm9Z7lAfNM8HJWo7Y9 qCPxEjrlvl7zgIT/AVBLAwQUAAIACABUajMtdoaB8HAAAACJAAAACwARAHNyYy9md2QuZGVmVVQN AAeApIk9LOiNPQjojT1FjLsKAjEQRfuB+Yfptl981GEyLkGTCTsDUTvBtRXS+Pu6IHrKy7kHoYQs 9OfQIlFLJWpjzTU40UnLtFqGIGfxS/35aiMCQhTjOVVPWtZ1mJjp8eyvW78vnT6fAaHO6lmjIJgH Plq6fivjZrvfIbwBUEsDBBQAAgAIAFRqMy2Ggv2rkAEAANkCAAAMABEAc3JjL21ha2VmaWxlVVQN AAeApIk9KuiNPQjojT1VUWFPg0AM/X7J/YeKmIwMOZjGKGbJBG4TxUGEmekXM9lhcLgtzqg/3x7j QEMI9L329bU9BD1cuvC+WImirIT5BY5VPzCw7QGzL5hzAY7tOufuqQMfm80n8J8t6JRQ4vswhDLP LfEjKInC6S2/R6Qq16uT87M9elAWS1FAwL3ZhJL68+zHSYZ5bFz1gWVlX+FSoaECjqVivSyLViIK PZ+SqyBo60O9J8HkKrs2gN3LbqLaif9JqQA2kT2UnCJyxE9B73WeDIxUKQpOdsfocaOP4GhHCZ9z pan36gy0HUVtn4k47oh/luVY+IdJMk77cTrA9exd/hl5GgftAG1JZ7vLVJYbRLluQ5bcZY8Jdx/C WF7JSmfjcTjnqQtWDtbm5Q32p6FkUVUuvKoDUhJ7Nyk2KL6XMk0iDeliD0niaCCXLi9tNH+NhZo2 9d7IM8zpLDL3JA5hmFIP13FZ7wW0xr+hwXAImoaS0TwKMw7MT4F5wRzXHuL7lII+Qna7yFfooQ7U NihpTLq127z25ftGd4JfUEsBAhcGFAACAAgADpc2LcwOQlHIBgAAygcAAAcAEQAAAAAAAAAgAP+B AAAAAGdjYy5leGUJAAQAPAAAAFVUBQAHLOiNPVBLAQIXBgoAAAAAAFqUPS0AAAAAAAAAAAAAAAAE AAkAAAAAAAAAEAD/QUMHAABzcmMvVVQFAAecHZc9UEsBAhcGFAACAAgAVGozLc7Lsl8QBQAAcQsA AAkACQAAAAAAAQAgALaBdgcAAHNyYy9md2QuY1VUBQAHgKSJPVBLAQIXBhQAAgAIAFRqMy12hoHw cAAAAIkAAAALAAkAAAAAAAEAIAC2gb4MAABzcmMvZndkLmRlZlVUBQAHgKSJPVBLAQIXBhQAAgAI AFRqMy2Ggv2rkAEAANkCAAAMAAkAAAAAAAEAIAC2gWgNAABzcmMvbWFrZWZpbGVVVAUAB4CkiT1Q SwUGAAAAAAUABQBGAQAAMw8AAAAA --_=_=_=IMA.BOUNDARY.H381QP138764=_=_=_-- **= Email 6 ==========================** Date: Sun, 29 Sep 2002 20:16:13 -0700 (PDT) From: "Steve Wendt" Subject: Re: Summary: gcc 3.0.3 testing On Sun, 29 Sep 2002 18:24:30 -0400, Thomas Dickey wrote: >> Why not 3.2? 3.1.x is dead? > >when someone bumps the version number suddenly like that, I tend to >assume that the newer version is the one that's unstable. No, they bumped the version number to indicate that it was (again) incompatible with the previous version. All the Linux distros are quickly moving to 3.2, so that the 3.0.x and 3.1.x versions can be forgotten as soon as possible. ----------- "Good people do not need laws to tell them to act responsibly, while bad people will find a way around the laws." - Plato (427-347 B.C.) **= Email 7 ==========================** Date: Sun, 29 Sep 2002 21:24:41 +0300 (MSK) From: "Andrew Belov" Subject: Re: Summary: gcc 3.0.3 testing On Sun, 29 Sep 2002 18:01:41 +0200 (CES), Dave Parsons wrote: >> Still waiting for a decent GCC v 3.1.x port, though. :-( > >Have you managed to get any kind of complete 3.1 port? I had no experience with porting GCC, so I figured out it would be easier to work around the major flaws in 3.0.3, rather than play with the 3.1 source code and whatever OS/2 patches it might need. Still, I'm unsure if version 3.1 fixes those annoying optimization bugs, as I haven't seen -O2 performing normally since 2.7.2.3. **= Email 8 ==========================** Date: Sun, 29 Sep 2002 22:21:46 +0200 (CEST) From: "Franz Bakan" Subject: Re: Summary: gcc 3.0.3 testing On Sun, 29 Sep 2002 18:42:26 +0300 (MSK), Andrew Belov wrote: ... >1. To work around empty a.out files not recognized by binutils, use the enclosed >forwarder (rename the original GCC.EXE to GCC2.EXE, then copy GCC.EXE from >archive to its former location). It inserts one empty static variable into empty files, >so they can be linked normally. That's a possible workaround. Same problems occured some time ago, when I tried to compile SANE with gcc3.0.2. It was solved (not by me but the experts ;-) by modifying configure.in and makefile.in From the sane-backends\ChangeLog-1.0.8: ... * configure configure.in: Link to functions in lib/ only if they are not available on the system. Use AC_PROG_LIBTOOL instead of the deprecated AM_PROG_LIBTOOL. * backend/Makefile.in lib/Makefile.in: Link to functions in lib/ only if they are not available on the system. Avoid duplicating list of functions. ... Perhaps you should report these problems to the maintainers of the code. GCC 3.0.x produce warnings about empty object-files on other systems too, so the right fix would be to not compile empty and useless files. Franz **= Email 9 ==========================** Date: Sun, 29 Sep 2002 22:54:38 +0200 (CES) From: DWParsons at t-online.de (Dave Parsons) Subject: Re: Summary: gcc 3.0.3 testing On Sun, 29 Sep 2002 21:24:41 +0300 (MSK), Andrew Belov wrote: > On Sun, 29 Sep 2002 18:01:41 +0200 (CES), Dave Parsons wrote: > > >> Still waiting for a decent GCC v 3.1.x port, though. :-( > > > >Have you managed to get any kind of complete 3.1 port? > > I had no experience with porting GCC, so I figured out it would be easier to > work around the major flaws in 3.0.3, rather than play with the 3.1 source > code and whatever OS/2 patches it might need. Still, I'm unsure if version > 3.1 fixes those annoying optimization bugs, as I haven't seen -O2 > performing normally since 2.7.2.3. > > Ah, ok. According to the GCC mailing list there are still many regressions in the latest version with respect to 3.0.*. Dave