From nobody@hyperreal.com Thu May 1 10:18:40 1997 Received: (from nobody@localhost) by hyperreal.com (8.8.5/8.8.5) id KAA04210; Thu, 1 May 1997 10:18:40 -0700 (PDT) Message-Id: <199705011718.KAA04210@hyperreal.com> Date: Thu, 1 May 1997 10:18:40 -0700 (PDT) From: Edward Tilley Reply-To: etilley@pentathalon.com To: apbugs@hyperreal.com Subject: Compile scripts fail on HPUX 10.20. X-Send-Pr-Version: 3.2 >Number: 520 >Category: os-hpux >Synopsis: Compile scripts fail on HPUX 10.20. >Confidential: no >Severity: serious >Priority: medium >Responsible: apache >State: closed >Class: sw-bug >Submitter-Id: apache >Arrival-Date: Thu May 1 10:20:01 1997 >Last-Modified: Wed May 7 08:26:42 PDT 1997 >Originator: etilley@pentathalon.com >Organization: >Release: 1.2b10 >Environment: HPUX 10.20 with barebones, vendor supplied ANSI-C compiler - Compiler was sufficient, however, to compile Perl 5.003 successfully so I so have confidence in it. >Description: I think I have found a bug in the most recent compile script. I can't compile Apache 1.2b10 on my HP-UX 10.20. During my "make" At line 544 of conf.h ... int ap_snprintf(char *buf, size_t len, const char *format,...); int ap_vsnprintf(char *buf, size_t len, const char *format, va_list ap); ... I get the following errors: (Bundled) cc: "conf.h", line 544: error 1000: Unexpected symbol: "size_t". (Bundled) cc: "conf.h", line 544: warning 5: "const" will become a keyword. (Bundled) cc: "conf.h", line 544: error 1000: Unexpected symbol: "char". (Bundled) cc: "conf.h", line 544: error 1000: Unexpected symbol: "...". (Bundled) cc: "conf.h", line 545: warning 5: "const" will become a keyword. (Bundled) cc: "conf.h", line 545: error 1000: Unexpected symbol: "const". (Bundled) cc: "conf.h", line 545: error 1000: Unexpected symbol: "va_list". (Bundled) cc: "/usr/include/sys/dirent.h", line 32: error 1000: Unexpected symbo l: "ino_t". (Bundled) cc: "conf.h", line 646: error 1000: Unexpected symbol: "Sigfunc". (Bundled) cc: "alloc.h", line 77: error 1000: Unexpected symbol: "*". (Bundled) cc: error 2017: Cannot recover from earlier errors, terminating. *** Error exit code 1 Here is a larger excerpt of conf.h surrounding the failing lines. #include #include /* * We use snprintf() to avoid overflows, but we include * our own version (ap_snprintf). Allow for people to use their * snprintf() if they want */ #ifdef HAVE_SNPRINTF #define ap_snprintf snprintf #define ap_vsnprintf vsnprintf #else int ap_snprintf(char *buf, size_t len, const char *format,...); int ap_vsnprintf(char *buf, size_t len, const char *format, va_list ap); #endif I am running everything in Configuration by the defaults except that my ANSI-C compiler does not support optimization (wierd but cheap). I don't program in C but I have reviewed the PORTING guide and still can't figure out whether this is a legitimate bug in the INSTALL procs. Also from conf.h, here are the HP defaults. #elif defined(HPUX) || defined(HPUX10) #define HAVE_SYS_RESOURCE_H #undef HAVE_GMTOFF #define NO_KILLPG #undef NO_SETSID #ifndef _HPUX_SOURCE #define _HPUX_SOURCE #endif #define JMP_BUF sigjmp_buf #define HAVE_SHMGET #ifndef HPUX10 #define SELECT_NEEDS_CAST typedef int rlim_t; #endif Any help would be greatly appreciated! >How-To-Repeat: Full description is pretty complete. email me at etilley@pentathalon.com if you want to try a quick fix and I will let you know how it ran on this platform. >Fix: Minor C src alterations required I think. If I knew C better, I might have been able to hack out a solution but I can't easily follow the assumptions the programmer made >Audit-Trail: From: coar@decus.org (Rodent of Unusual Size) To: apbugs@apache.org, etilley@pentathalon.com, Coar@decus.org Subject: Re: os-hpux/520: Compile scripts fail on HPUX 10.20. Date: Sun, 04 May 1997 09:08:39 -0400 [This information was added as a separate problem report; I am adding it to the text of PR#520, where it belongs.] From: drz@apache.att.com I have played around with this release and try to make it build on HP-UX. I had the same problem with what you submitted initially. However, what I found out is there is something wroug with the script that can't generate the right set of CFLAGS. To be brief, here is what you can do: Modify the Makefile after you run the Configure and set CFLAGS as: CFLAGS1= -Aa -w -DHPUX -D_HPUX_SOURCE or CFLAGS1= -Aa -v -w -DHPUX10 -D_HPUX_SOURCE If you use HP/UX 10. It should make it build. David -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ David Zhuang Eml: drz@apache.mt.att.com Web: http://apache.mt.att.com:8000/~drz ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ From: coar@decus.org (Rodent of Unusual Size) To: apbugs@apache.org, etilley@pentathalon.com, Coar@decus.org Subject: Re: os-hpux/520: Compile scripts fail on HPUX 10.20. Date: Sun, 04 May 1997 09:20:04 -0400 [Another addition from improperly sent mail] BTW: No -O2 flag, but you can set +O2 if you like. David -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ David Zhuang Eml: drz@apache.mt.att.com Web: http://apache.mt.att.com:8000/~drz ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ From: coar@decus.org (Rodent of Unusual Size) To: apbugs@apache.org, etilley@pentathalon.com, Coar@decus.org Subject: Re: os-hpux/520: Compile scripts fail on HPUX 10.20. Date: Mon, 05 May 1997 20:25:22 -0400 [Additional information from the original submitter. Please note that *I* (Ken) am not the author of the reply; is. I just attached his notes to the problem report.] From the fingers of Edward Tilley flowed the following: > > Hi, > >Thanks for your response. Unfortunately, I did not purchase the HPUX >addon C compiler for this system. Strange, because as I mentioned in my >bug report, that the barebones compiler that I do have was sufficient to >compile PERL 5.003. > >-Aa nor +O2 are supported and the Apache install script will not run >whether I use your recommended changes or not. > >Thanks just the same. > >Cheers, > >Ed State-Changed-From-To: open-closed State-Changed-By: marc State-Changed-When: Wed May 7 08:26:41 PDT 1997 State-Changed-Why: User reports they are using HPUX's free compiler (which is _NOT_ ANSI). Although Larry may be able to avoid using ANSI features Apache does and isn't going to change. The best solution is gcc. >Unformatted: