On Sep 22, 2014, at 6:19 AM, Mrunal Nerpawar <p.mru...@gmail.com> wrote:
> > > On Mon, Sep 22, 2014 at 3:45 PM, Mrunal Nerpawar <p.mru...@gmail.com> wrote: > > > On Sat, Sep 20, 2014 at 7:28 AM, Tom Francis <fr...@tcsaf.com> wrote: > > On Sep 19, 2014, at 7:55 AM, Mrunal Nerpawar <p.mru...@gmail.com> wrote: > > > > > > > On Fri, Sep 19, 2014 at 3:05 PM, Andy Polyakov <ap...@openssl.org> wrote: > > > I could not get this working even on a 11.23 machine having latest (last > > > one released in December 2007). > > > > ... latest what? I mean it appears that you missed something. But more > > relevant question is what does ld -V return, what patches are installed. > > But please don't show the list and ask which patches should be > > installed. Or at least don't expect answer from me, as I don't know. As > > already mentioned, the references to specific ld versions is simply > > something that was observed to not work and then work in specific > > situation, it's just two dots on the mesh, that's all. > > > > I meant latest patch.. I ate that word > > There are definitely more recent patches than 2007. HP fixed a number of > bugs (one that I submitted) for Itanium and 11.23 in 2011; took a few tries > to get it right, unfortunately. Bugs in compiler, link editor, and dynamic > linker. As Andy mentions, the output of ld -V can at least give us an idea > of what you’ve got. I’ve no longer got access to the HP boxes I was using, > so I can only go by memory here. :( Oh, and I still get patch notifications > from HP, so I know they’re still making new patches for 11.23 on a regular > basis. :) > > Okay. I will check for the patches. It is admin guy who has the creds to > download the patches install them on servers. > output of ld cmd: > bash-2.05# ld -V > ld: 92453-07 linker ld HP Itanium(R) B.12.42 IPF/IPF > ld: No input files > Fatal error. > > bash-2.05# cc -V > cc: HP C/aC++ B3910B A.06.25 [Nov 30 2009] For cc, that’s good. That’s the version I was using. :) That ld version is much too old — I’m surprised you can link anything — you’re supposed to have a newer version of the link editor in order install that version of the compiler (but you can get around that, which I’m guessing your admin did :( ). The admin needs to install ALL of the associated patches for that compiler, which means patches for ld, dld, as, and a lot more. It’s a huge set of patches, at least several hundred MB. Somebody put together this page: http://hpux.connect.org.uk/hppd/answers/4-4.html that seems to list several of the relevant patches, and has links to download them. Please note, that these patches may be a little out of date, but the good news is, the admin who follows those links will see links to more recent patches if they exist. :) > > > Can anyone tell me the impact of not using +sectionmerge since it is not > > > working on 11.23. > > > > It was mentioned it the beginning of thread: "it didn't do the > > sectionmerge (*which was resulting in crash upon startup*)". It should > > be noted that crash is one possibility, another possibility is that > > initialization test is rendered *non-compliant* with validation > > requirements. > > Right. I shouldn't have asked this question again. Apologies. > > > > > I am also evaluating the possibility of using 11.31 and tried building > > > openssl on it, however, getting different error while bulding fips code > > > this time: > > > > > > + cc -I. -I.. -I../include -DOPENSSL_FIPSCANISTER +Z -DOPENSSL_PIC > > > -DOPENSSL_THREADS -DDSO_DLFCN -DHAVE_DLFCN_H -D_REENTRANT -Ae +DD64 +O3 > > > (Bundled) cc: warning 922: > > ^^^^^^^^^ Real, i.e. unbundled, compiler is absolute requirement for > > OpenSSL on HP-UX in any situation. > > Okay. Let me try getting unbundled compiler and try compiling > > > > > Another build issue on HP-PA: > > > I am getting following error on HP-PA while building fips capable openssl: > > > > > > make[2]: Entering directory `/openssl/hp-pa/openssl-1.0.1h' > > > [ -z "libcrypto" ] || cc +Z -DOPENSSL_PIC -DOPENSSL_THREADS -DDSO_DL > > > -D_REENTRANT +DA2.0 +DS2.0 +O3 +Optrs_strongly_typed -Ae +ESlit > > > -DB_ENDIAN -DMD32_XARRAY -D_REENTRANT -I/openssl/hp-pa/fips/include > > > -Iinclude \ > > > -DFINGERPRINT_PREMAIN_DSO_LOAD -o fips_premain_dso \ > > > /openssl/hp-pa/fips/lib/fips_premain.c > > > /openssl/hp-pa/fips/lib/fipscanister.o \ > > > libcrypto.a -Wl,+s -ldld > > > /usr/ccs/bin/ld: Duplicate symbol "$global$" in files > > > /opt/langtools/lib/crt0.o and /openssl/hp-pa/fips/lib/fipscanister.o > > > > I don't have comment on this, because I don't have access to HP-PA > > system with vendor compiler. I can only say that goal is that > > fipscanister.o looks exactly like any other .o file generated by > > compiler, and just like any other such file it shouldn't have symbols > > otherwise found in crt0.o. The fact that crt0.o symbols made their way > > to fipscanister.o indicates that something went wrong at fipscanister.o > > link stage and -r flag was not respected. Is it possible that it also > > was linked with bundled compiler? > > It is not. Here is the output: > > bash-2.05$ which cc > > /bin/cc > > bash-2.05$ what /bin/cc > > /bin/cc: > > $Revision: 92453-07 linker linker crt0.o B.11.64 080728 $ > > LINT B.11.11.22 CXREF B.11.11.22 > > HP92453-01 B.11.11.22 HP C Compiler > > $ PATCH/11.00:PHCO_27774 Oct 3 2002 09:45:59 $ > > what is just showing the version info of libc and crt0, which while helpful > aren’t relevant. cc -V will give the version number. But that’s OK, because > I remember this problem, and it’s most annoying, every so often, an object > file is generated with symbols from crt0.o included — made a real mess of > compiling any application; you’d get random failures at link time with > different object files reporting the multiple symbols. The fix here is a bit > different, though, you need to go download the new HP C/C++ compiler package > and install it — you won’t be offered patches that will fix the problem (the > latest patch for that compiler version is the one that introduced the > problem). Be aware, that in order to install the newer compiler, you’ll need > to install a lot of other patches — if you’re in the developer program, it > should offer to let you download all the relevant patches at once (at least, > it did back in 2011, which was the last time I had to run a similar upgrade; > HP stopped offering new versions as patches around 2009, IIRC). > > One other thing: How are you checking if you’ve got the latest patches > installed? > > TOM > > cc -V isn't showing anything. That means it’s either the bundled C compiler or it’s pretty old (prior to 6.x ?). Either way, I recommend at least 6.20 for OpenSSL and 6.25 if you want to do C++. I’ve used both successfully with OpenSSL, but it’s possible that 6.20 was only used for PA-RISC, not for Itanium. > As mentioned above, its the admin guy who installed and informed that it is > at latest now. > I could not verify as swlist command is also throwing error, > swlist -l patches > ERROR: An invalid level was given for the successful execution of > "swlist". The valid levels are: "root", "depot", "bundle", > "product", "subproduct", "fileset", "file", "shroot", > "prroot", "category", "patch", "control_file". > ERROR: Command line parsing failed. > It’s 'swlist -l patch’ But if you don’t have access to the HP patch site, that list won’t do you much good, except to maybe help prove how out of date your system is. :) > Btw, are these error while building openssl seen from certain version of > openssl or so? > The last version built on these machines was 0.9.8r around 3 yrs back and no > such issues were seen then. The compiler issues I referenced with regards to crt0.o were not from OpenSSL specifically, but from my employer’s application (and just about anything else I wanted to compile — and no, the problem didn’t always occur, but it was frequent). One of HP’s patches (related to the compiler) introduced the problem, and it couldn’t be resolved without also updating the compiler. The patches that go with the newer compiler will fix the problem (and installing the patches without the newer compiler still leaves it broken — I never looked into why). Did you compile that version of OpenSSL? Or was it installed by the admin? If the admin, it might’ve been a binary package from HP. :) > Mrunal > > > > > ______________________________________________________________________ > > OpenSSL Project http://www.openssl.org > > Development Mailing List openssl-dev@openssl.org > > Automated List Manager majord...@openssl.org > > > ______________________________________________________________________ > OpenSSL Project http://www.openssl.org > Development Mailing List openssl-dev@openssl.org > Automated List Manager majord...@openssl.org > > ______________________________________________________________________ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
