Archive-Date: Wed, 01 Jan 1997 04:08:55 +0200 Sender: owner-emacs-testers@lp.se MIME-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Date: Wed, 1 Jan 1997 04:08:51 MET Message-ID: <13001.58195.271759.feedmail.pl5.LEVITTE@devil.bofh.se> From: Richard Levitte - VMS Whacker Reply-To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU Subject: Re: gcc can't find config.h? From: Roland Roberts I'm confused as to why config.h isn't found---it does exist in the src directory. How does mms/mmk do subprocesses? Is it possible the logical isn't getting set in the environment used by gcc? MMS and MMK start one subprocess, to which it feeds all the actions of a desciptor file. I suspect gcc doesn't handle searchlists well, which is a pity... Pat Rankin could answer to that question... -- R Levitte, Levitte Programming; Spannv. 38, I; S-161 43 Bromma; SWEDEN Tel: +46-8-26 52 47; Cel: +46-10-222 64 05; No fax right now PGP key fingerprint = A6 96 C0 34 3A 96 AA 6C B0 D5 9A DF D2 E9 9C 65 Check http://www.lp.se/~levitte for my public key. bastard@bofh.se ================================================================================ Archive-Date: Wed, 01 Jan 1997 04:17:18 +0200 Sender: owner-emacs-testers@lp.se MIME-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Date: Wed, 1 Jan 1997 04:17:10 MET Message-ID: <13001.58694.271759.feedmail.pl5.LEVITTE@devil.bofh.se> From: Richard Levitte - VMS Whacker Reply-To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU Subject: Re: Some initial gcc change requirements From: Roland Roberts Additionally, I've tweaked gcc's include path to find gnu_cc_include:[vms] before sys$library. This is necessary to avoid finding all those nasty VAXC'isms like "variant_struct". BTW, Pat Rankin's gcc install setup does a very nice job of converting all the VMS headers to usable form. Thanks for the info, I'll make sure autoconf gets it right. sys_siglist test fails), but all of the library tests are failing. I suspect I'm not linking against gcclib and thus the consistent failure. The linking problem is probably due to incorrect declaration of the GCC librarirs. I'll attend to that this evening. checking for sys_siglist declaration in signal.h or unistd.h conftest.c: In function `t': conftest.c:8: `sys_siglist' undeclared (first use this function) conftest.c:8: (Each undeclared identifier is reported only once conftest.c:8: for each function it appears in.) %GCC-E-NOOBJECT, no object file was produced Damn interesting! SYS$ERROR is redirected to NL:, and the configuration script has done the following: SET MESSAGE/NOFAC/NOSEV/NOID/NOTEXT And still, you get those messages. Makes you wonder what gcc is doing... creating [.src]descrip.mms junk.c:1256: invalid preprocessing directive name Interesting. Please try to find out what happened... -- R Levitte, Levitte Programming; Spannv. 38, I; S-161 43 Bromma; SWEDEN Tel: +46-8-26 52 47; Cel: +46-10-222 64 05; No fax right now PGP key fingerprint = A6 96 C0 34 3A 96 AA 6C B0 D5 9A DF D2 E9 9C 65 Check http://www.lp.se/~levitte for my public key. bastard@bofh.se ================================================================================ Archive-Date: Wed, 01 Jan 1997 12:12:34 +0200 Sender: owner-emacs-testers@lp.se MIME-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Date: Wed, 1 Jan 1997 12:12:28 MET Message-ID: <13002.21676.271759.feedmail.pl5.LEVITTE@devil.bofh.se> From: Richard Levitte - VMS Whacker Reply-To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU Subject: Re: Some initial gcc change requirements From: Roland Roberts Additionally, I've tweaked gcc's include path to find gnu_cc_include:[vms] before sys$library. This is necessary to avoid finding all those nasty VAXC'isms like "variant_struct". BTW, Pat Rankin's gcc install setup does a very nice job of converting all the VMS headers to usable form. I'll admit I don't understand. First of all, according to what I can see in GCC_STARTUP.COM, gnu_cc_include is defined as a searchlist which includes the vms subdirectory. Therefore, gnu_cc_include: should be suffisient (sp?). Second, when I read GCC.HLP, I find the following: /INCLUDE_DIRECTORY /INCLUDE_DIRECTORY=(path[,path...]) ... There are two forms for specifying include files - #include "file-spec" and #include . For the #include "file-spec" form, the search order is: 1. The directory containing the source file. 2. The directories in the /INCLUDE qualifier (if any). 3. The directory (or directories) specified in the logical name GNU_CC_INCLUDE. 4. The directory (or directories) specified in the logical name SYS$LIBRARY. For the #include form, the search order is: 1. The directories specified in the /INCLUDE qualifier (if any). 2. The directory (or directories) specified in the logical name GNU_CC_INCLUDE. 3. The directory (or directories) specified in the logical name SYS$LIBRARY. As you can see, gnu_cc_include is used before sys$library by default. Or could it be that you're talkig about the definition if sys_includes? -- R Levitte, Levitte Programming; Spannv. 38, I; S-161 43 Bromma; SWEDEN Tel: +46-8-26 52 47; Cel: +46-10-222 64 05; No fax right now PGP key fingerprint = A6 96 C0 34 3A 96 AA 6C B0 D5 9A DF D2 E9 9C 65 Check http://www.lp.se/~levitte for my public key. bastard@bofh.se ================================================================================ Archive-Date: Wed, 01 Jan 1997 12:23:45 +0200 Sender: owner-emacs-testers@lp.se MIME-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Date: Wed, 1 Jan 1997 12:23:40 MET Message-ID: <13002.22348.271759.feedmail.pl5.LEVITTE@devil.bofh.se> From: Richard Levitte - VMS Whacker Reply-To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU Subject: Re: Some initial gcc change requirements Roland, I've just made the small change to configure.com that puts gnu_cc_include at the beginning of the definition of sys_includes. Please fetch the new configure.com from ftp://vms.gnu.ai.mit.edu/gnu-vms/software/snapshots/emacs-19_28/ and try it out. Ah, also, I've patched [.src.m]vax.h to make sure GCCLIB.OLB is used as well as VAXCRTL, when GNU C is used. The patched version is available in ftp://vms.gnu.ai.mit.edu/gnu-vms/software/snapshots/emacs-19_28/src/m/ (damn, those paths get long...) -- R Levitte, Levitte Programming; Spannv. 38, I; S-161 43 Bromma; SWEDEN Tel: +46-8-26 52 47; Cel: +46-10-222 64 05; No fax right now PGP key fingerprint = A6 96 C0 34 3A 96 AA 6C B0 D5 9A DF D2 E9 9C 65 Check http://www.lp.se/~levitte for my public key. bastard@bofh.se ================================================================================ Archive-Date: Wed, 01 Jan 1997 23:15:34 +0200 Sender: owner-emacs-testers@lp.se Date: Wed, 1 Jan 1997 17:15:23 -0500 (EST) Message-ID: <199701012215.RAA06871@galileo.> From: "Roland B. Roberts" To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU Subject: Re: Some initial gcc change requirements Reply-To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU >>>>> "Richard" == Richard Levitte <- VMS Whacker > writes: Richard> I've just made the small change to configure.com that puts Richard> gnu_cc_include at the beginning of the definition of Richard> sys_includes. Please fetch the new configure.com from Sorry for the confusion...that shouldn't be necessary. What I meant by my original was that the default setup for gcc is probably incorrect. It sets up a default gnu_cc_inlude which includes all the "standard" headers but puts all of the VMS specific headers in gnu_cc_include:[vms] which means you would have to include them via, for example, #include Since that is ugly, and because I am writing code using Wollongong TCP/IP, I've redefined gnu_cc_include to be a search list (which gcc handles just fine). BTW, gnu_cc_include needs to look like a conceal device name. So I have a logical $ define gcc$include disk1:[dirA.dirB.gcc.include.], - disk1:[dirA.dirB.gcc.include.vms.] $ define twg$include disk2:[dirC.twg.include.] $ define gnu_cc_include gcc$include, twg$include Notice I can get away without actually making them concealed devices (yes, it really works). I really think the first definition ought to be the default for gnu_cc_include in order to pick up the converted VMS headers. Without it, gcc looks in sys$library. Richard> Ah, also, I've patched [.src.m]vax.h to make sure Richard> GCCLIB.OLB is used as well as VAXCRTL, when GNU C is used. Thanks! I'll grab it tomorrow when I'm back at work roland -- Roland B Roberts roberts@panix.com ================================================================================ Archive-Date: Wed, 01 Jan 1997 23:17:43 +0200 Sender: owner-emacs-testers@lp.se Date: Wed, 1 Jan 1997 17:17:33 -0500 (EST) Message-ID: <199701012217.RAA06890@galileo.> From: "Roland B. Roberts" To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU CC: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU Subject: Re: Some initial gcc change requirements Reply-To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU >>>>> "Richard" == Richard Levitte <- VMS Whacker > writes: Richard> I'll admit I don't understand. First of all, according to Richard> what I can see in GCC_STARTUP.COM, gnu_cc_include is Richard> defined as a searchlist which includes the vms Richard> subdirectory. Therefore, gnu_cc_include: should be Richard> suffisient (sp?). Second, when I read GCC.HLP, I find the Richard> following: I'll have to look again...for some reason, this wasn't what was happening for me...maybe I broke something (quite possible...I hacked in my logical for Wollongong quite quickly and probably have glossed over checking for multiple translations of gnu_cc_include when I added the path to Wollongong's headers). roland -- Roland B Roberts roberts@panix.com ================================================================================ Archive-Date: Wed, 01 Jan 1997 23:20:27 +0200 Sender: owner-emacs-testers@lp.se Date: Wed, 1 Jan 1997 17:20:19 -0500 (EST) Message-ID: <199701012220.RAA06916@galileo.> From: "Roland B. Roberts" To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU Subject: Re: Some initial gcc change requirements Reply-To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU checking for sys_siglist declaration in signal.h or unistd.h conftest.c: In function `t': conftest.c:8: `sys_siglist' undeclared (first use this function) conftest.c:8: (Each undeclared identifier is reported only once conftest.c:8: for each function it appears in.) %GCC-E-NOOBJECT, no object file was produced Richard> Damn interesting! SYS$ERROR is redirected to NL:, and the Richard> configuration script has done the following: Richard> SET MESSAGE/NOFAC/NOSEV/NOID/NOTEXT Richard> And still, you get those messages. Makes you wonder what Richard> gcc is doing... I wonder if it is something happening in gcc's DCL glue to run the pieces. Or maybe it is spitting something to stdout instead of stderr.... I'll take a look if I can. creating [.src]descrip.mms junk.c:1256: invalid preprocessing directive name Richard> Interesting. Please try to find out what happened... Will do... roland -- Roland B Roberts roberts@panix.com ================================================================================ Archive-Date: Thu, 02 Jan 1997 00:40:39 +0200 Sender: owner-emacs-testers@lp.se MIME-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Date: Thu, 2 Jan 1997 00:40:35 MET Message-ID: <13003.1027.271759.feedmail.pl5.LEVITTE@devil.bofh.se> From: Richard Levitte - VMS Whacker Reply-To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU To: emacs-testers@LISTS.VMS.GNU.AI.MIT.EDU Subject: Re: Some initial gcc change requirements From: "Roland B. Roberts" >>>>> "Richard" == Richard Levitte <- VMS Whacker > writes: Richard> I've just made the small change to configure.com that puts Richard> gnu_cc_include at the beginning of the definition of Richard> sys_includes. Please fetch the new configure.com from Sorry for the confusion...that shouldn't be necessary. I'm not too sure... I'll play around a little... What I meant by my original was that the default setup for gcc is probably incorrect. It sets up a default gnu_cc_inlude which includes all the "standard" headers but puts all of the VMS specific headers in gnu_cc_include:[vms] which means you would have to include them via, for example, #include That's pretty strange. I just did the following: $ @pd:[src.gcc-vms-2_7_1.gcc]gcc_startup $ sh log gnu_cc_include "GNU_CC_INCLUDE" = "DISK$VAXVMS061:[PD.SRC.GCC-VMS-2_7_1.GCC.INCLUDE.]" (LNM$SYSTEM_TABLE) = "DISK$VAXVMS061:[PD.SRC.GCC-VMS-2_7_1.GCC.INCLUDE.VMS.]" Now, seeing the second value, I wonder why I need to include instead of . Or does your gcc_startup.com look different from mine? I'll just say at once: I don't recall where I got my copy of GCC. It was months ago, but I haven't had the time to start playing with it until today. -- R Levitte, Levitte Programming; Spannv. 38, I; S-161 43 Bromma; SWEDEN Tel: +46-8-26 52 47; Cel: +46-10-222 64 05; No fax right now PGP key fingerprint = A6 96 C0 34 3A 96 AA 6C B0 D5 9A DF D2 E9 9C 65 Check http://www.lp.se/~levitte for my public key. bastard@bofh.se