Permissions on installed shared libraries
In Makefile the permissions for shared libaries are set to 644. However, HP-UX (32bit) atleast requires then to be set to executable for dld to load them. Will it do any harm to have then install 755 for all OSes? Cheers, Ross @if [ -n $(SHARED_LIBS) ]; then \ tmp=$(SHARED_LIBS); \ for i in $${tmp:-x}; \ do \ if [ -f $$i -o -f $$i.a ]; then \ ( echo installing $$i; \ if [ $(PLATFORM) != Cygwin ]; then \ cp $$i $(INSTALL_PREFIX) $(INSTALLTOP)/lib; \ chmod 755 $(INSTALL_PREFIX) $(INSTALLTOP)/lib/$$i; \ else \ c=`echo $$i | sed 's/^lib/cyg/'`; \ cp $$c $(INSTALL_PREFIX) $(INSTALLTOP)/bin/$$c; \ chmod 644 $(INSTALL_PREFIX) $(INSTALLTOP)/bin/$$c; \ cp $$i.a $(INSTALL_PREFIX) $(INSTALLTOP)/lib/$$i.a; \ chmod 644 $(INSTALL_PREFIX) $(INSTALLTOP)/lib/$$i.a; \ fi ); \ fi; \ done; \ - Ross Alexander He knows no more about his MIS - NEC Europe Limiteddestiny than a tea leaf knows Work ph: +44 20 8752 3394 the history of East India Company __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: [openssl.org #96] bug in config script (gcc 3.1)
Lutz, I will create a patch against openssl-0.9.7-STABLE-snap-020612 with the gcc change and the __LP__ check for 64bit HPUX. If I get time I may look at dynamic linking. Cheers, Ross - Ross Alexander He knows no more about his MIS - NEC Europe Limiteddestiny than a tea leaf knows Work ph: +44 20 8752 3394 the history of East India Company |-+- | | Lutz Jaenicke via| | | RT | | | [EMAIL PROTECTED] | | | Sent by: | | | owner-openssl-dev@| | | openssl.org | | | | | | | | | 13/06/2002 08:09 | | | Please respond to | | | openssl-dev | | | | |-+- -| | | | To: [EMAIL PROTECTED], [EMAIL PROTECTED] | | cc: [EMAIL PROTECTED] | | Subject: [openssl.org #96] bug in config script (gcc 3.1) | -| [[EMAIL PROTECTED] - Thu Jun 13 08:34:54 2002]: The config script needs to use gcc -dumpversion instead of gcc --version to determine the gcc version. gcc-3.1 outputs a bunch of text with --version, but just the number with -dumpversion, which also works for gcc-2.95. Sounds like a possible solution to the problem of gcc version recognition. The solution being included currently doesn't seem to be sufficient (see thread Various patches for 0.9.6d and 0.9.7-beta1, I have added [EMAIL PROTECTED] as another requestor for this ticket). For how long this flag has been supported? I suppose it will work on CYGWIN as well. Best regards, Lutz __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: [openssl.org #96] bug in config script (gcc 3.1)
Lutz, Here is the patch for config. I'm not sure if the 32bit/64bit gcc test is really in the right place, but its not too bad. (See attached file: config.diff) Cheers, Ross - Ross Alexander He knows no more about his MIS - NEC Europe Limiteddestiny than a tea leaf knows Work ph: +44 20 8752 3394 the history of East India Company |-+- | | Lutz Jaenicke via| | | RT | | | [EMAIL PROTECTED] | | | Sent by: | | | owner-openssl-dev@| | | openssl.org | | | | | | | | | 13/06/2002 08:09 | | | Please respond to | | | openssl-dev | | | | |-+- -| | | | To: [EMAIL PROTECTED], [EMAIL PROTECTED] | | cc: [EMAIL PROTECTED] | | Subject: [openssl.org #96] bug in config script (gcc 3.1) | -| [[EMAIL PROTECTED] - Thu Jun 13 08:34:54 2002]: The config script needs to use gcc -dumpversion instead of gcc --version to determine the gcc version. gcc-3.1 outputs a bunch of text with --version, but just the number with -dumpversion, which also works for gcc-2.95. Sounds like a possible solution to the problem of gcc version recognition. The solution being included currently doesn't seem to be sufficient (see thread Various patches for 0.9.6d and 0.9.7-beta1, I have added [EMAIL PROTECTED] as another requestor for this ticket). For how long this flag has been supported? I suppose it will work on CYGWIN as well. Best regards, Lutz __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED] config.diff Description: Binary data
Various patches for 0.9.6d and 0.9.7-beta1
Hello, I have two patches for hpux64-parisc-gcc support in Configure and one to correct an error in evp_test.c, which calls strsep instead of sstrsep (09.7-beta1 only). (See attached file: openssl-0.9.7-beta1.patch)(See attached file: openssl-0.9.6d.patch)(See attached file: openssl-0.9.7-beta1.evp_test.patch) I have tests the hpux64 with gcc-3.1.0 and make test runs through correctly. These are not the most optimal values (eg it sets the -g flag and no optimization) but gcc for hppa64-hp-hpux11.00 is still a little flaky. This currently doesn't create shared libraries. The GNU linker is very flaky but I wouldn't take too much to create shared libaries with the HP linker. Cheers Ross - Ross Alexander He knows no more about his MIS - NEC Europe Limiteddestiny than a tea leaf knows Work ph: +44 20 8752 3394 the history of East India Company openssl-0.9.7-beta1.patch Description: Binary data openssl-0.9.6d.patch Description: Binary data openssl-0.9.7-beta1.evp_test.patch Description: Binary data
Re: Various patches for 0.9.6d and 0.9.7-beta1
Lutz, I think the problem is because 32bit HP uses the SOM object format and 64bit HP uses ELF. These are quite different and hence gcc cannot be configured for both targets. So when you build gcc it is a different target (hppa64-hp-hpux11.00) for 64bit than 32bit (hppa2.0n-hp-hpux11.00). NB On PA-linux this is different, its uses ELF32 and ELF64 respectively, so I think a single instance of gcc does have 32/64 bit flag. Thats why there are no flags to tell gcc to produce 64bit code, it will always operate in LP64 mode. Because there is no real autoconfigure you have to assume the user knows what they are doing. I was mainly using openssl to test gcc, as it has a very solid test regime. I hope this explaination makes sense. Cheers, Ross - Ross Alexander He knows no more about his MIS - NEC Europe Limiteddestiny than a tea leaf knows Work ph: +44 20 8752 3394 the history of East India Company |-+- | | Lutz Jaenicke | | | [EMAIL PROTECTED]| | | Cottbus.DE | | | Sent by: | | | owner-openssl-dev@open| | | ssl.org | | | | | | | | | 10/06/2002 14:27 | | | Please respond to | | | openssl-dev | | | | |-+- -| | | | To: [EMAIL PROTECTED] | | cc: | | Subject: Re: Various patches for 0.9.6d and 0.9.7-beta1 | -| On Mon, Jun 10, 2002 at 01:52:12PM +0100, [EMAIL PROTECTED] wrote: I have two patches for hpux64-parisc-gcc support in Configure and one to correct an error in evp_test.c, which calls strsep instead of sstrsep (09.7-beta1 only). Thanks. The strsep issue is already fixed in the current tree. (See attached file: openssl-0.9.7-beta1.patch)(See attached file: openssl-0.9.6d.patch)(See attached file: openssl-0.9.7-beta1.evp_test.patch) I have tests the hpux64 with gcc-3.1.0 and make test runs through correctly. These are not the most optimal values (eg it sets the -g flag and no optimization) but gcc for hppa64-hp-hpux11.00 is still a little flaky. This currently doesn't create shared libraries. The GNU linker is very flaky but I wouldn't take too much to create shared libaries with the HP linker. I don't yet fully understand your patches (I only have gcc-3.0.x on HP-UX 10.20, so I am not familiar with the latest 64bit issues). It however seems, that nowhere any 64bit command line option is used. Doesn't this mean, that 32bit code is generated? For hpux64-parisc-cc the +DD64 flag is required. Best regards, Lutz -- Lutz Jaenicke [EMAIL PROTECTED] http://www.aet.TU-Cottbus.DE/personen/jaenicke/ BTU Cottbus, Allgemeine Elektrotechnik Universitaetsplatz 3-4, D-03044 Cottbus __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: Various patches for 0.9.6d and 0.9.7-beta1
So the entries you supplied are for gcc (hppa64-hp-hpux11.00)? Is there a way for config to find out itself? Please have a look into config and search for GCC_ARCH to see what I mean. Sure. It will take me a couple of days. In GCC 3.1 gcc --version doesn't work the same way so I will looking at gcc -v | egrep ^gcc version to do the same job. It may be best to compile a program like the following ... #include stdio.h int main() { printf(%d.%d.%d\n, __GNUC__, __GNUC_MINOR__, __GNUC_PATCHLEVEL__); return 0; } This way you have complete control over the format. Cheers, Ross - Ross Alexander He knows no more about his MIS - NEC Europe Limiteddestiny than a tea leaf knows Work ph: +44 20 8752 3394 the history of East India Company |-+- | | Lutz Jaenicke | | | [EMAIL PROTECTED]| | | Cottbus.DE | | | Sent by: | | | owner-openssl-dev@open| | | ssl.org | | | | | | | | | 10/06/2002 15:14 | | | Please respond to | | | openssl-dev | | | | |-+- -| | | | To: [EMAIL PROTECTED] | | cc: | | Subject: Re: Various patches for 0.9.6d and 0.9.7-beta1 | -| On Mon, Jun 10, 2002 at 03:06:22PM +0100, [EMAIL PROTECTED] wrote: I think the problem is because 32bit HP uses the SOM object format and 64bit HP uses ELF. These are quite different and hence gcc cannot be configured for both targets. So when you build gcc it is a different target (hppa64-hp-hpux11.00) for 64bit than 32bit (hppa2.0n-hp-hpux11.00). NB On PA-linux this is different, its uses ELF32 and ELF64 respectively, so I think a single instance of gcc does have 32/64 bit flag. So the entries you supplied are for gcc (hppa64-hp-hpux11.00)? Is there a way for config to find out itself? Please have a look into config and search for GCC_ARCH to see what I mean. Best regards, Lutz -- Lutz Jaenicke [EMAIL PROTECTED] http://www.aet.TU-Cottbus.DE/personen/jaenicke/ BTU Cottbus, Allgemeine Elektrotechnik Universitaetsplatz 3-4, D-03044 Cottbus __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: Various patches for 0.9.6d and 0.9.7-beta1
Lutz, I've picked up the snapshot. There is a problem with the new code. GCCVER=`(gcc --version) 2/dev/null | head -1` if [ $GCCVER != ]; then CC=gcc # then strip off whatever prefix Cygnus as well as GCC 3.1 prepends # the number with... Hopefully, this will work for any future prefixes # as well. GCCVER=`echo $GCCVER | sed 's/^[a-zA-Z ()]*\-//'` # peak single digit before and after first dot, e.g. 2.95.1 gives 29 GCCVER=`echo $GCCVER | sed 's/\([0-9]\)\.\([0-9]\).*/\1\2/'` else CC=cc fi GCCVER=${GCCVER:-0} The problem is with GCCVER=`echo $GCCVER | sed 's/^[a-zA-Z ()]*\-//'` The \- doesn't work with gcc 3.1 which outputs gcc (GCC) 3.1. I can fix the regex but I like know the reasoning behind it. I tried ^[a-zA-Z ()]*\-? but this doesn't work with sed because it an extended regex and not supported by HP or GNU sed. As for detecting 64bit GCC, that looks fairly easy. GCC passes a define -D__LP64__ to cpp. Otherwise it would something like int main() { if ((sizeof(long) == 8) (sizeof(void*) == 8)) return 1; /* LP64 */ or ((sizeof(long == 8) || (sizeof(void*) == 8)) return -1; /* error, this should never happen on hpux */ or return 0; } would work. Any suggestions? Ross - Ross Alexander He knows no more about his MIS - NEC Europe Limiteddestiny than a tea leaf knows Work ph: +44 20 8752 3394 the history of East India Company |-+- | | Lutz Jaenicke | | | [EMAIL PROTECTED]| | | Cottbus.DE | | | Sent by: | | | owner-openssl-dev@open| | | ssl.org | | | | | | | | | 10/06/2002 15:46 | | | Please respond to | | | openssl-dev | | | | |-+- -| | | | To: [EMAIL PROTECTED] | | cc: | | Subject: Re: Various patches for 0.9.6d and 0.9.7-beta1 | -| On Mon, Jun 10, 2002 at 03:44:07PM +0100, [EMAIL PROTECTED] wrote: So the entries you supplied are for gcc (hppa64-hp-hpux11.00)? Is there a way for config to find out itself? Please have a look into config and search for GCC_ARCH to see what I mean. Sure. It will take me a couple of days. In GCC 3.1 gcc --version doesn't work the same way so I will looking at gcc -v | egrep ^gcc version to do the same job. Please load down a current snapshot. GCC-3.1 support for --version should be added. Best regards, Lutz -- Lutz Jaenicke [EMAIL PROTECTED] http://www.aet.TU-Cottbus.DE/personen/jaenicke/ BTU Cottbus, Allgemeine Elektrotechnik Universitaetsplatz 3-4, D-03044 Cottbus __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]