Re: M2Crypto 0.18 - new version, same old build bugs - more details
John Nagle wrote: > Heikki Toivonen wrote: >That's progress, but the build still doesn't work: ... > during a C compile, we get > > SWIG/_m2crypto_wrap.c:2529:18: error: _lib.h: No such file or directory > > and the build goes downhill from there, with many compile errors in the > GCC phase. The gcc call > > gcc -pthread -fno-strict-aliasing -DNDEBUG -g -O3 -Wall -Wstrict-prototypes > -fPIC -I/usr/local/include/python2.5 -c SWIG/_m2crypto_wrap.c -o > build/temp.linux-i686-2.5/SWIG/_m2crypto_wrap.o -DTHREADING > > is in the directory below SWIG, and doesn't include SWIG, so it's clear > why gcc couldn't find the file. Any sign of a fix yet? John Nagle -- http://mail.python.org/mailman/listinfo/python-list
Re: M2Crypto 0.18 - new version, same old build bugs - more details
Heikki Toivonen wrote: > John Nagle wrote: > >>But notice that the -D entry didn't appear on the SWIG command line. >>Neither did the "-includeall". The "swig_opts" values around line >>129 aren't actually being used. I think that's left over from the code >>intended >>to allow builds with Python 2.3 and earlier. The "self.swig_opts" up at >>line 53 of "setup.py" seems to be controlling. The patch was to >>the obsolete code. > > > Aha! Good find. I reopened > https://bugzilla.osafoundation.org/show_bug.cgi?id=9404 and attached a > patch that should address this for real this time. At least -includeall > appears in my Ubuntu Dapper Drake environment. Could you give it a go > and let me know how it works? That's progress, but the build still doesn't work: $ python setup.py build running build running build_py running build_ext building 'M2Crypto.__m2crypto' extension swigging SWIG/_m2crypto.i to SWIG/_m2crypto_wrap.c swig -python -I/usr/local/include/python2.5 -I/usr/include -includeall -D__i386__ -o SWIG/_m2crypto_wrap.c SWIG/_m2crypto.i /usr/include/openssl/opensslconf.h:13: Error: Unable to find 'opensslconf-i386.h' error: command 'swig' failed with exit status 1 Now that SWIG is being invoked with "-includeall", it has to have all the include paths defined. (Without "includeall", includes are deferred until the C compile stage.) Note that the SWIG command line specifies "/usr/include", but not "/usr/include/openssl". Within M2Crypto's SWIG files, there's _ec.i:%include which brings in "opensslconf.h" from /usr/include/openssl. But that file has #if defined(__i386__) #include "opensslconf-i386.h" #elif defined(__ia64__) ... Since "opensslconf-i386.h" lives in /usr/include/openssl, and that directory isn't mentioned on the SWIG command line, the #include of "opensslconf-i386.h" fails. And, no, adding self.swig_opts.append('-DOPENSSL_NO_EC') # Uncomment if you can't build with EC disabled does not help. As a test, I tried adding "-I/usr/include/openssl" to the SWIG command line. Building then gets further, but during a C compile, we get SWIG/_m2crypto_wrap.c:2529:18: error: _lib.h: No such file or directory and the build goes downhill from there, with many compile errors in the GCC phase. The gcc call gcc -pthread -fno-strict-aliasing -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -fPIC -I/usr/local/include/python2.5 -c SWIG/_m2crypto_wrap.c -o build/temp.linux-i686-2.5/SWIG/_m2crypto_wrap.o -DTHREADING is in the directory below SWIG, and doesn't include SWIG, so it's clear why gcc couldn't find the file. So those fixes were not enough. Include file management in the M2Crypto build clearly needs some work. John Nagle -- http://mail.python.org/mailman/listinfo/python-list
Re: M2Crypto 0.18 - new version, same old build bugs - more details
John Nagle wrote: > But notice that the -D entry didn't appear on the SWIG command line. > Neither did the "-includeall". The "swig_opts" values around line > 129 aren't actually being used. I think that's left over from the code > intended > to allow builds with Python 2.3 and earlier. The "self.swig_opts" up at > line 53 of "setup.py" seems to be controlling. The patch was to > the obsolete code. Aha! Good find. I reopened https://bugzilla.osafoundation.org/show_bug.cgi?id=9404 and attached a patch that should address this for real this time. At least -includeall appears in my Ubuntu Dapper Drake environment. Could you give it a go and let me know how it works? -- Heikki Toivonen -- http://mail.python.org/mailman/listinfo/python-list
Re: M2Crypto 0.18 - new version, same old build bugs - more details
Heikki Toivonen wrote: > John Nagle wrote: > >>Back in March, I posted this: >> >> >>>Hit that with OpenSSL. Red Hat took elliptical curve cryptography >>>out of Fedora 6 for patent reasons. With that missing, M2Crypto won't >>>build. It ought to; the implementor of M2Crypto thought of that, because >>>it's an optional feature. But M2Crypto uses SWIG. SWIG doesn't normally >>>process nested include files. OpenSSL had changed their configuration >>>approach to use nested include files. So SWIG didn't see the #define >>>that >>>said to turn off elliptical curve crypto support. This resulted in >>>compile errors in the 24,000 lines of code that come out of SWIG. >> >> That was for M2Crypto 0.17. >> >> It's still broken in M2Crypto 0.18. > > > This was reported as > https://bugzilla.osafoundation.org/show_bug.cgi?id=9404 and fixed, at > least according to the person who reported the bug. I'd be curious to > know why the fix does not work for you... > > As an alternative, does it work if you add line 130 to setup.py: > > '-DOPENSSL_NO_EC', # Uncomment if you have > trouble compiling without EC support OK, here's the build failing: python setup.py build running build running build_py running build_ext building 'M2Crypto.__m2crypto' extension swigging SWIG/_m2crypto.i to SWIG/_m2crypto_wrap.c swig -python -I/usr/local/include/python2.5 -I/usr/include -o SWIG/_m2crypto_wrap.c SWIG/_m2crypto.i /usr/include/openssl/opensslconf.h:27: Error: CPP #error ""This openssl-devel package does not work your architecture?"". Use the -cpperraswarn option to continue swig processing. error: command 'swig' failed with exit status 1 OK, let's check the requirements, from http://chandlerproject.org/bin/view/Projects/MeTooCrypto M2Crypto 0.18: * Python 2.3 or newer o m2urllib2 requires Python 2.4 or newer * OpenSSL 0.9.7 or newer o Some optional new features will require OpenSSL 0.9.8 or newer * SWIG 1.3.24 or newer python --version Python 2.5 [OK here, 2.5 >= 2.4] swig -version SWIG Version 1.3.31 [OK here; 1.3.31 >= 1.3.24] Compiled with i386-redhat-linux-g++ [i386-redhat-linux-gnu] openssl OpenSSL> version OpenSSL 0.9.8a 11 Oct 2005 [OK here; 0.9.8a => 0.9.7.] Tried the suggested patch to "setup.py": diff setup.py.orig setup.py 130c130 < #'-D__i386__', # Uncomment for early OpenSSL 0.9.7 versions --- > '-D__i386__', # Uncomment for early > OpenSSL 0.9.7 versions No change in error message. That gets us -bash-3.1$ python setup.py build running build running build_py running build_ext building 'M2Crypto.__m2crypto' extension swigging SWIG/_m2crypto.i to SWIG/_m2crypto_wrap.c swig -python -I/usr/local/include/python2.5 -I/usr/include -o SWIG/_m2crypto_wrap.c SWIG/_m2crypto.i /usr/include/openssl/opensslconf.h:27: Error: CPP #error ""This openssl-devel package does not work your architecture?"". Use the -cpperraswarn option to continue swig processing. error: command 'swig' failed with exit status 1 But notice that the -D entry didn't appear on the SWIG command line. Neither did the "-includeall". The "swig_opts" values around line 129 aren't actually being used. I think that's left over from the code intended to allow builds with Python 2.3 and earlier. The "self.swig_opts" up at line 53 of "setup.py" seems to be controlling. The patch was to the obsolete code. This is on Fedora Core 5, x86, 32 bit. John Nagle -- http://mail.python.org/mailman/listinfo/python-list
Re: M2Crypto 0.18 - new version, same old build bugs.
John Nagle wrote: > Back in March, I posted this: > >> Hit that with OpenSSL. Red Hat took elliptical curve cryptography >> out of Fedora 6 for patent reasons. With that missing, M2Crypto won't >> build. It ought to; the implementor of M2Crypto thought of that, because >> it's an optional feature. But M2Crypto uses SWIG. SWIG doesn't normally >> process nested include files. OpenSSL had changed their configuration >> approach to use nested include files. So SWIG didn't see the #define >> that >> said to turn off elliptical curve crypto support. This resulted in >> compile errors in the 24,000 lines of code that come out of SWIG. > >That was for M2Crypto 0.17. > >It's still broken in M2Crypto 0.18. This was reported as https://bugzilla.osafoundation.org/show_bug.cgi?id=9404 and fixed, at least according to the person who reported the bug. I'd be curious to know why the fix does not work for you... As an alternative, does it work if you add line 130 to setup.py: '-DOPENSSL_NO_EC', # Uncomment if you have trouble compiling without EC support -- Heikki Toivonen -- http://mail.python.org/mailman/listinfo/python-list
Re: M2Crypto 0.18 - new version, same old build bugs.
Paul Boddie wrote: > On 20 Sep, 22:12, John Nagle <[EMAIL PROTECTED]> wrote: > >>That was for M2Crypto 0.17. >> >>It's still broken in M2Crypto 0.18. >> >>And there's no RPM or Windows binary. >> >>Nobody actually uses this stuff, do they? > > > I can find Ubuntu packages for 0.16 and 0.17: > > http://packages.ubuntu.com/feisty/libs/python-m2crypto > http://packages.ubuntu.com/gutsy/libs/python-m2crypto > > So, clearly some people are managing to build the library, even if it > isn't the absolute latest version. You might want to take a look at > the source packages and the patches the distribution people apply, > although it doesn't look like they are actually doing much patching. > The source packages suggest that they use SWIG 1.3.24 to build > M2Crypto - is that what you're using? > > Paul I have builds of 0.17 of M2Crypto for both Linux and Windows. I want the new bug fixes. The 0.17 version mishandles SSL certs that specify more than one domain name, and that's supposedly fixed in 0.18. But there are enough changes that just copying over the relevant file (Checker.py) isn't sufficient. John Nagle -- http://mail.python.org/mailman/listinfo/python-list
Re: M2Crypto 0.18 - new version, same old build bugs.
On 20 Sep, 22:12, John Nagle <[EMAIL PROTECTED]> wrote: > > That was for M2Crypto 0.17. > > It's still broken in M2Crypto 0.18. > > And there's no RPM or Windows binary. > > Nobody actually uses this stuff, do they? I can find Ubuntu packages for 0.16 and 0.17: http://packages.ubuntu.com/feisty/libs/python-m2crypto http://packages.ubuntu.com/gutsy/libs/python-m2crypto So, clearly some people are managing to build the library, even if it isn't the absolute latest version. You might want to take a look at the source packages and the patches the distribution people apply, although it doesn't look like they are actually doing much patching. The source packages suggest that they use SWIG 1.3.24 to build M2Crypto - is that what you're using? Paul -- http://mail.python.org/mailman/listinfo/python-list
M2Crypto 0.18 - new version, same old build bugs.
Back in March, I posted this: > Hit that with OpenSSL. Red Hat took elliptical curve cryptography > out of Fedora 6 for patent reasons. With that missing, M2Crypto won't > build. It ought to; the implementor of M2Crypto thought of that, because > it's an optional feature. But M2Crypto uses SWIG. SWIG doesn't normally > process nested include files. OpenSSL had changed their configuration > approach to use nested include files. So SWIG didn't see the #define that > said to turn off elliptical curve crypto support. This resulted in > compile errors in the 24,000 lines of code that come out of SWIG. That was for M2Crypto 0.17. It's still broken in M2Crypto 0.18. And there's no RPM or Windows binary. Nobody actually uses this stuff, do they? John Nagle -- http://mail.python.org/mailman/listinfo/python-list