openvpn-devel Mailing List for OpenVPN

Again, All conditional statements should be within single config.h or config-msvc.h or sysheader.h or anything. Doing conditional inline makes code complex and unmaintainable. But as you adds loooooooong patches without proper review I guess it is not that important. Alon. On Thu, Feb 16, 2012 at 8:47 PM, David Sommerseth <da...@re...> wrote: > From: Heiko Hund <hei...@so...> > > MSVC does not know wstat(). Instead _wstat() must be used here. > Unfortunately _wstat() takes a 'struct _stat'. A type 'stat_t' is > introduced to handle this situation in a portable way. > > [v2: Use openvpn_stat_t instead of stat_t (David Sommerseth)] > > Signed-off-by: Heiko Hund <hei...@so...> > Signed-off-by: David Sommerseth <da...@re...> > --- >  misc.h  |    6 ++++-- >  pf.c    |    2 +- >  win32.c |    4 ++-- >  3 files changed, 7 insertions(+), 5 deletions(-) > > diff --git a/misc.h b/misc.h > index e7a0b55..bdada42 100644 > --- a/misc.h > +++ b/misc.h > @@ -168,10 +168,12 @@ openvpn_open (const char *path, int flags, mode_t mode) >  #endif > >  #ifdef WIN32 > -int openvpn_stat (const char *path, struct stat *buf); > +typedef struct _stat openvpn_stat_t; > +int openvpn_stat (const char *path, openvpn_stat_t *buf); >  #else > +typedef struct stat openvpn_stat_t; >  static inline int > -openvpn_stat (const char *path, struct stat *buf) > +openvpn_stat (const char *path, openvpn_stat_t *buf) >  { >   return stat (path, buf); >  } > diff --git a/pf.c b/pf.c > index a0e9fc8..0ef839e 100644 > --- a/pf.c > +++ b/pf.c > @@ -498,7 +498,7 @@ pf_check_reload (struct context *c) >       && c->c2.pf.filename >       && event_timeout_trigger (&c->c2.pf.reload, &c->c2.timeval, ETT_DEFAULT)) >     { > -      struct stat s; > +      openvpn_stat_t s; >       if (!openvpn_stat (c->c2.pf.filename, &s)) >        { >          if (s.st_mtime > c->c2.pf.file_last_mod) > diff --git a/win32.c b/win32.c > index a8f4ed9..2ba97fc 100644 > --- a/win32.c > +++ b/win32.c > @@ -1073,10 +1073,10 @@ openvpn_open (const char *path, int flags, int mode) >  } > >  int > -openvpn_stat (const char *path, struct stat *buf) > +openvpn_stat (const char *path, openvpn_stat_t *buf) >  { >   struct gc_arena gc = gc_new (); > -  int res = wstat (wide_string (path, &gc), buf); > +  int res = _wstat (wide_string (path, &gc), buf); >   gc_free (&gc); >   return res; >  } > -- > 1.7.4.4 > > > ------------------------------------------------------------------------------ > Virtualization & Cloud Management Using Capacity Planning > Cloud computing makes use of virtualization - but cloud computing > also focuses on allowing computing to be delivered as a service. > http://www.accelacomm.com/jaw/sfnl/114/51521223/ > _______________________________________________ > Openvpn-devel mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/openvpn-devel 

Hi, On Thu, Feb 16, 2012 at 07:47:25PM +0100, David Sommerseth wrote: > From: Heiko Hund <hei...@so...> > > MSVC does not know wstat(). Instead _wstat() must be used here. > Unfortunately _wstat() takes a 'struct _stat'. A type 'stat_t' is > introduced to handle this situation in a portable way. > > [v2: Use openvpn_stat_t instead of stat_t (David Sommerseth)] > > Signed-off-by: Heiko Hund <hei...@so...> > Signed-off-by: David Sommerseth <da...@re...> ACK! gert -- USENET is *not* the non-clickable part of WWW! //www.muc.de/~gert/ Gert Doering - Munich, Germany ge...@gr... fax: +49-89-35655025 ge...@ne... 

From: Heiko Hund <hei...@so...> MSVC does not know wstat(). Instead _wstat() must be used here. Unfortunately _wstat() takes a 'struct _stat'. A type 'stat_t' is introduced to handle this situation in a portable way. [v2: Use openvpn_stat_t instead of stat_t (David Sommerseth)] Signed-off-by: Heiko Hund <hei...@so...> Signed-off-by: David Sommerseth <da...@re...> --- misc.h | 6 ++++-- pf.c | 2 +- win32.c | 4 ++-- 3 files changed, 7 insertions(+), 5 deletions(-) diff --git a/misc.h b/misc.h index e7a0b55..bdada42 100644 --- a/misc.h +++ b/misc.h @@ -168,10 +168,12 @@ openvpn_open (const char *path, int flags, mode_t mode) #endif #ifdef WIN32 -int openvpn_stat (const char *path, struct stat *buf); +typedef struct _stat openvpn_stat_t; +int openvpn_stat (const char *path, openvpn_stat_t *buf); #else +typedef struct stat openvpn_stat_t; static inline int -openvpn_stat (const char *path, struct stat *buf) +openvpn_stat (const char *path, openvpn_stat_t *buf) { return stat (path, buf); } diff --git a/pf.c b/pf.c index a0e9fc8..0ef839e 100644 --- a/pf.c +++ b/pf.c @@ -498,7 +498,7 @@ pf_check_reload (struct context *c) && c->c2.pf.filename && event_timeout_trigger (&c->c2.pf.reload, &c->c2.timeval, ETT_DEFAULT)) { - struct stat s; + openvpn_stat_t s; if (!openvpn_stat (c->c2.pf.filename, &s))	{ if (s.st_mtime > c->c2.pf.file_last_mod) diff --git a/win32.c b/win32.c index a8f4ed9..2ba97fc 100644 --- a/win32.c +++ b/win32.c @@ -1073,10 +1073,10 @@ openvpn_open (const char *path, int flags, int mode) } int -openvpn_stat (const char *path, struct stat *buf) +openvpn_stat (const char *path, openvpn_stat_t *buf) { struct gc_arena gc = gc_new (); - int res = wstat (wide_string (path, &gc), buf); + int res = _wstat (wide_string (path, &gc), buf); gc_free (&gc); return res; } -- 1.7.4.4 

On Thu, Feb 16, 2012 at 7:30 PM, Heiko Hund <hei...@so...> wrote: > Windows API CommandLineToArgvW(), introduced in Windows unicode path > commit 71bbbd76c62630c88441237d72fe5b61f0b45b2a, is defined therein. Usually this should be avoided and get command-line from wmain(). 

I stopped following all OpenVPN changes. It seems like more changes are entered than should without proper review. Anyway, this is not the correct solution. Correct solution is to have config-msvc.h and have: --- #define mode_t int --- And in autoconf (if mingw does not have this as well): AC_CHECK_DECLS( [mode_t], , [AC_DEFINE([mode_t],[int], [Emulate mode_t])], [[ #include <fcntl.h> ]] ) On Thu, Feb 16, 2012 at 7:30 PM, Heiko Hund <hei...@so...> wrote: > > The MSVC headers do not define mode_t. open() uses an int for > the permissions instead. Fixes building with the MSVC based > buildsystem. > > Signed-off-by: Heiko Hund <hei...@so...> > --- >  misc.h  |    2 +- >  win32.c |    2 +- >  2 files changed, 2 insertions(+), 2 deletions(-) > > diff --git a/misc.h b/misc.h > index 8c0bae1..e7a0b55 100644 > --- a/misc.h > +++ b/misc.h > @@ -158,7 +158,7 @@ openvpn_fopen (const char *path, const char *mode) >  #endif > >  #ifdef WIN32 > -int openvpn_open (const char *path, int flags, mode_t mode); > +int openvpn_open (const char *path, int flags, int mode); >  #else >  static inline int >  openvpn_open (const char *path, int flags, mode_t mode) > diff --git a/win32.c b/win32.c > index 5b38918..a8f4ed9 100644 > --- a/win32.c > +++ b/win32.c > @@ -1064,7 +1064,7 @@ openvpn_fopen (const char *path, const char *mode) >  } > >  int > -openvpn_open (const char *path, int flags, mode_t mode) > +openvpn_open (const char *path, int flags, int mode) >  { >   struct gc_arena gc = gc_new (); >   int fd = _wopen (wide_string (path, &gc), flags, mode); > -- > 1.7.5.4 > > > ------------------------------------------------------------------------------ > Virtualization & Cloud Management Using Capacity Planning > Cloud computing makes use of virtualization - but cloud computing > also focuses on allowing computing to be delivered as a service. > http://www.accelacomm.com/jaw/sfnl/114/51521223/ > _______________________________________________ > Openvpn-devel mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/openvpn-devel 

Hi, On Thu, Feb 16, 2012 at 06:30:40PM +0100, Heiko Hund wrote: > MSVC does not know wstat(). Instead _wstat() must be used here. > Unfortunately _wstat() takes a 'struct _stat'. A type 'stat_t' is > introduced to handle this situation in a portable way. Semi-ACK. In general, ACK, though it might be a bit easier to understand the code if there were an explantory comment at the typedef, and the type would be called "openvpn_stat_t" or such - so that when referenced in pf.c, it's clear that this is not something "from the build environment" but "local". gert -- USENET is *not* the non-clickable part of WWW! //www.muc.de/~gert/ Gert Doering - Munich, Germany ge...@gr... fax: +49-89-35655025 ge...@ne... 

Hi, On Thu, Feb 16, 2012 at 06:30:39PM +0100, Heiko Hund wrote: > The MSVC headers do not define mode_t. open() uses an int for > the permissions instead. Fixes building with the MSVC based > buildsystem. ACK. (I was slightly worried that this might break mingw building, but David assures me that you've tested this :-) ). gert -- USENET is *not* the non-clickable part of WWW! //www.muc.de/~gert/ Gert Doering - Munich, Germany ge...@gr... fax: +49-89-35655025 ge...@ne... 

Hi, On Thu, Feb 16, 2012 at 06:30:38PM +0100, Heiko Hund wrote: > Use of TARGET_WIN32 breaks MSVC builds as it is only defined > for mingw builds done with the autotools buildsystem. ACK. I think it's reasonable to have exactly one CPP symbol to identify "building in the windows 32bit environment". I have no particular preference, and there are less TARGET_WIN32s in the code - so get rid of that one :-) (Yes, we should get rid of all the windows build cruft, but the focus for now is "have something that *builds* and can be tagged as 2.3-alpha1"). gert -- USENET is *not* the non-clickable part of WWW! //www.muc.de/~gert/ Gert Doering - Munich, Germany ge...@gr... fax: +49-89-35655025 ge...@ne... 

Hi, On Thu, Feb 16, 2012 at 06:30:41PM +0100, Heiko Hund wrote: > Windows API CommandLineToArgvW(), introduced in Windows unicode path > commit 71bbbd76c62630c88441237d72fe5b61f0b45b2a, is defined therein. > > Signed-off-by: Heiko Hund <hei...@so...> Since you've tested it and it makes Samuli happy -> ACK :-) gert -- USENET is *not* the non-clickable part of WWW! //www.muc.de/~gert/ Gert Doering - Munich, Germany ge...@gr... fax: +49-89-35655025 ge...@ne... 

Windows API CommandLineToArgvW(), introduced in Windows unicode path commit 71bbbd76c62630c88441237d72fe5b61f0b45b2a, is defined therein. Signed-off-by: Heiko Hund <hei...@so...> --- win/msvc.mak.in | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/win/msvc.mak.in b/win/msvc.mak.in index 115e395..191f370 100644 --- a/win/msvc.mak.in +++ b/win/msvc.mak.in @@ -31,7 +31,7 @@ LZO_DYNAMIC = lzo2.lib INCLUDE_DIRS = -I$(OPENSSL)/include -I$(POLARSSL)/include -I$(LZO)/include -I$(PKCS11_HELPER)/include -LIBS = $(OPENSSL_DYNAMIC) $(POLARSSL_DYNAMIC) $(PKCS11_HELPER_DYNAMIC) $(LZO_DYNAMIC) ws2_32.lib crypt32.lib iphlpapi.lib winmm.lib user32.lib gdi32.lib advapi32.lib wininet.lib +LIBS = $(OPENSSL_DYNAMIC) $(POLARSSL_DYNAMIC) $(PKCS11_HELPER_DYNAMIC) $(LZO_DYNAMIC) ws2_32.lib crypt32.lib iphlpapi.lib winmm.lib user32.lib gdi32.lib advapi32.lib wininet.lib shell32.lib LIB_DIRS = -LIBPATH:$(OPENSSL)\lib -LIBPATH:$(POLARSSL)\build\library -LIBPATH:$(PKCS11_HELPER)\lib -LIBPATH:$(LZO)\lib -- 1.7.5.4 

The MSVC headers do not define mode_t. open() uses an int for the permissions instead. Fixes building with the MSVC based buildsystem. Signed-off-by: Heiko Hund <hei...@so...> --- misc.h | 2 +- win32.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/misc.h b/misc.h index 8c0bae1..e7a0b55 100644 --- a/misc.h +++ b/misc.h @@ -158,7 +158,7 @@ openvpn_fopen (const char *path, const char *mode) #endif #ifdef WIN32 -int openvpn_open (const char *path, int flags, mode_t mode); +int openvpn_open (const char *path, int flags, int mode); #else static inline int openvpn_open (const char *path, int flags, mode_t mode) diff --git a/win32.c b/win32.c index 5b38918..a8f4ed9 100644 --- a/win32.c +++ b/win32.c @@ -1064,7 +1064,7 @@ openvpn_fopen (const char *path, const char *mode) } int -openvpn_open (const char *path, int flags, mode_t mode) +openvpn_open (const char *path, int flags, int mode) { struct gc_arena gc = gc_new (); int fd = _wopen (wide_string (path, &gc), flags, mode); -- 1.7.5.4 

MSVC does not know wstat(). Instead _wstat() must be used here. Unfortunately _wstat() takes a 'struct _stat'. A type 'stat_t' is introduced to handle this situation in a portable way. Signed-off-by: Heiko Hund <hei...@so...> --- misc.h | 6 ++++-- pf.c | 2 +- win32.c | 4 ++-- 3 files changed, 7 insertions(+), 5 deletions(-) diff --git a/misc.h b/misc.h index e7a0b55..55d494f 100644 --- a/misc.h +++ b/misc.h @@ -168,10 +168,12 @@ openvpn_open (const char *path, int flags, mode_t mode) #endif #ifdef WIN32 -int openvpn_stat (const char *path, struct stat *buf); +typedef struct _stat stat_t; +int openvpn_stat (const char *path, stat_t *buf); #else +typedef struct stat stat_t; static inline int -openvpn_stat (const char *path, struct stat *buf) +openvpn_stat (const char *path, stat_t *buf) { return stat (path, buf); } diff --git a/pf.c b/pf.c index a0e9fc8..67cd28f 100644 --- a/pf.c +++ b/pf.c @@ -498,7 +498,7 @@ pf_check_reload (struct context *c) && c->c2.pf.filename && event_timeout_trigger (&c->c2.pf.reload, &c->c2.timeval, ETT_DEFAULT)) { - struct stat s; + stat_t s; if (!openvpn_stat (c->c2.pf.filename, &s))	{ if (s.st_mtime > c->c2.pf.file_last_mod) diff --git a/win32.c b/win32.c index a8f4ed9..4efc8df 100644 --- a/win32.c +++ b/win32.c @@ -1073,10 +1073,10 @@ openvpn_open (const char *path, int flags, int mode) } int -openvpn_stat (const char *path, struct stat *buf) +openvpn_stat (const char *path, stat_t *buf) { struct gc_arena gc = gc_new (); - int res = wstat (wide_string (path, &gc), buf); + int res = _wstat (wide_string (path, &gc), buf); gc_free (&gc); return res; } -- 1.7.5.4 

Use of TARGET_WIN32 breaks MSVC builds as it is only defined for mingw builds done with the autotools buildsystem. Signed-off-by: Heiko Hund <hei...@so...> --- misc.c | 6 +++--- misc.h | 6 +++--- openvpn.c | 2 +- 3 files changed, 7 insertions(+), 7 deletions(-) diff --git a/misc.c b/misc.c index 58096fa..a07780f 100644 --- a/misc.c +++ b/misc.c @@ -356,7 +356,7 @@ int openvpn_chdir (const char* dir) { #ifdef HAVE_CHDIR -#ifdef TARGET_WIN32 +#ifdef WIN32 int res; struct gc_arena gc = gc_new (); res = _wchdir (wide_string (dir, &gc)); @@ -598,7 +598,7 @@ openvpn_system (const char *command, const struct env_set *es, unsigned int flag /* * execute the command */ -#ifdef TARGET_WIN32 +#ifdef WIN32 gc = gc_new (); ret = _wsystem (wide_string (command, &gc)); gc_free (&gc); @@ -627,7 +627,7 @@ openvpn_system (const char *command, const struct env_set *es, unsigned int flag int openvpn_access (const char *path, int mode) { -#ifdef TARGET_WIN32 +#ifdef WIN32 struct gc_arena gc = gc_new (); int ret = _waccess (wide_string (path, &gc), mode); gc_free (&gc); diff --git a/misc.h b/misc.h index 2413f53..8c0bae1 100644 --- a/misc.h +++ b/misc.h @@ -147,7 +147,7 @@ openvpn_run_script (const struct argv *a, const struct env_set *es, const unsign return openvpn_execve_check(a, es, flags | S_SCRIPT, msg); }; -#ifdef TARGET_WIN32 +#ifdef WIN32 FILE * openvpn_fopen (const char *path, const char *mode); #else static inline FILE * @@ -157,7 +157,7 @@ openvpn_fopen (const char *path, const char *mode) } #endif -#ifdef TARGET_WIN32 +#ifdef WIN32 int openvpn_open (const char *path, int flags, mode_t mode); #else static inline int @@ -167,7 +167,7 @@ openvpn_open (const char *path, int flags, mode_t mode) } #endif -#ifdef TARGET_WIN32 +#ifdef WIN32 int openvpn_stat (const char *path, struct stat *buf); #else static inline int diff --git a/openvpn.c b/openvpn.c index 84289d2..b2175a1 100644 --- a/openvpn.c +++ b/openvpn.c @@ -131,7 +131,7 @@ main (int argc, char *argv[]) return 1; #endif -#ifdef TARGET_WIN32 +#ifdef WIN32 SetConsoleOutputCP (CP_UTF8); #endif -- 1.7.5.4 

Hi, this patch series fixes several issues when building for Windows with the MSVC buildsystem discovered today: [PATCH 1/4] replace check for TARGET_WIN32 with WIN32 [PATCH 2/4] do not use mode_t on Windows [PATCH 3/4] use the underscore version of stat on Windows [PATCH 4/4] make MSVC link against shell32 as well For details check the commit messages of the individual patches. Regards Heiko -- Heiko Hund | Software Engineer | Phone +49-721-25516-237 | Fax -200 Astaro a Sophos Company | Amalienbadstr. 41 Bau 52 | 76227 Karlsruhe | Germany Commercial Register: Mannheim HRA 702710 | Headquarter Location: Karlsruhe Represented by the General Partner Astaro Verwaltungs GmbH Amalienbadstraße 41 Bau 52 | 76227 Karlsruhe | Germany Commercial Register: Mannheim HRB 708248 | Executive Board: Gert Hansen, Markus Hennig, Jan Hichert, Günter Junk, Dr. Frank Nellissen 

Frank de Brabander wrote: > Below I have included a patch that changes the bool return type to an int. > This will result in 'make check' to pass. I see that you already committed this. % cd openvpn % make distclean % git pull % git checkout 6449a149f850e9e1207233f3ca642d9342fbfbaf % autoreconf -i -v % ./configure --with-lzo-headers=/opt/local/include \ --with-lzo-lib=/opt/local/lib % make % make check (wait for it...) ==================== All 2 tests passed (1 test was not run) ==================== Works like a charm! Thanks, both of you! Freek 

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 16/02/12 15:00, Frank de Brabander wrote: > Sorry, patch format got broken in previous email. Hope this works > better. No worries, patches is tricky in quite some MUAs. Anyhow, patch is applied to master on -testing and -stable trees. commit 6449a149f850e9e1207233f3ca642d9342fbfbaf Author: Frank de Brabander <bra...@fo...> Date: Thu Feb 16 13:51:04 2012 +0100 Fixed wrong return type of cipher_kt_mode The cipher_kt_mode uses bool as return type, this should be int. On some platforms like OS X, any returned value larger than one will be converted to 1. Signed-off-by: Frank de Brabander <bra...@fo...> Acked-by: Adriaan de Jong <de...@fo...> Signed-off-by: David Sommerseth <da...@re...> Kind regards, David Sommerseth -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAk89EC4ACgkQDC186MBRfrpLNgCcDo+3ZRWi5T62iUZVEjaVOsBH +cIAoKQbItKZqgR+f372PCqCy8Fuo3ia =pey+ -----END PGP SIGNATURE----- 

Sorry, patch format got broken in previous email. Hope this works better. 

 -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Jun, I would try David's obfsproxy + OpenVPN suggestion first and reconsider if that fails. If we would add obfuscation into OpenVPN itself, we'd soon be in the same cat and mouse game as Tor, with the exception we don't have the same amount of developer interest to follow it through. So, I think keeping the obfuscation functionality would be challenging in the long run. - -- Samuli Seppänen Community Manager OpenVPN Technologies, Inc irc freenode net: mattock > Hi David, > > I replied to you directly, maybe it got caught in your spam filter? > > Anyone else has some thoughts about developing an obfsproxy-style > component for OpenVPN? > > Best, > > Jun. > > On 13/02/2012 13:09, David Sommerseth wrote: > > On 13/02/12 13:59, Jun Matsushita wrote: > >> This is my first post in this list. As probably a lot of you > >> heard, Iran has stepped up its filtering by apparently blocking > >> SSL/TLS using DPI. This is a good read about what's happening > >> http://news.ycombinator.com/item?id=3575029. As these statistics > >> from TOR attest > >> https://metrics.torproject.org/users.html?graph=direct-users&start=2012-01-12&end=2012-02-12&country=ir&events=on&dpi=72#direct-users > > >> > > > the impact has been immediate and surely concerns the majority of > > tools > >> out there. > > >> Does OpenVPN allow the use of some form of Obfuscation (such as > >> the one TOR is testing now and seem to work from within Iran > >> https://www.torproject.org/projects/obfsproxy-instructions.html.en)? > > >> Anyone has thoughts about this or would be interested in > >> discussing the matter? > > > > Hi Jun, > > > OpenVPN uses SSL under the hood. But it does some tricks to allow > > SSL over UDP (SSL is strictly designed for TCP). This however > > makes some changes that many DPI firewalls *might* not identify > > OpenVPN traffic as SSL traffic. But as it's not really an > > obfuscation which changes over time, it might still be possible to > > block it if this kind of traffic is detected. > > > However, the obfsproxy project sounds very interesting. And it > > should be possible to use obfsproxy (as it can talk like a SOCKS > > proxy) with OpenVPN, by using the --socks-proxy argument. But I'm > > not aware of any openvpn services providing obfsproxy services in > > conjunction with OpenVPN. > > > > kind regards, > > > David Sommerseth > > ------------------------------------------------------------------------------ > Try before you buy = See our experts in action! > The most comprehensive online learning library for Microsoft developers > is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3, > Metro Style Apps, more. Free future releases when you subscribe now! > http://p.sf.net/sfu/learndevnow-dev2 > _______________________________________________ > Openvpn-devel mailing list > Ope...@li... > https://lists.sourceforge.net/lists/listinfo/openvpn-devel -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAk88/isACgkQwp2X7RmNIqMv1ACfXT8xpTLVNng2RsVSukurCWEt gMQAoNxqob4FGdi60axCsSbz1Q6SVGOH =l436 -----END PGP SIGNATURE----- 

> -----Original Message----- > From: Frank de Brabander [mailto:bra...@fo...] > > Hello, > > I have looked into the problem, it seems to be caused by the return > type of cipher_kt_mode() in crypto_backend.h being declared as a bool. > This function is called from init_key_type() of crypto.c during the > tests. The actual mode value in cipher_kt_mode() of openssl_backend.c > during the failed test is 2, but is converted to 1 due to this boolean > return type. > My guess is that this is only happening on OS X because of a specific > GCC compiler version or custom Apple patches. > > Below I have included a patch that changes the bool return type to an > int. > This will result in 'make check' to pass. > Ack, I saw this work on Frank's mac :). Apologies for the missed bool. Adriaan 

Hello, I have looked into the problem, it seems to be caused by the return type of cipher_kt_mode() in crypto_backend.h being declared as a bool. This function is called from init_key_type() of crypto.c during the tests. The actual mode value in cipher_kt_mode() of openssl_backend.c during the failed test is 2, but is converted to 1 due to this boolean return type. My guess is that this is only happening on OS X because of a specific GCC compiler version or custom Apple patches. Below I have included a patch that changes the bool return type to an int. This will result in 'make check' to pass. diff --git a/crypto_backend.h b/crypto_backend.h index 8622962..a0966dd 100644 --- a/crypto_backend.h +++ b/crypto_backend.h @@ -218,7 +218,7 @@ int cipher_kt_block_size (const cipher_kt_t *cipher_kt); * @return Cipher mode, either \c OPENVPN_MODE_CBC, \c * OPENVPN_MODE_OFB or \c OPENVPN_MODE_CFB */ -bool cipher_kt_mode (const cipher_kt_t *cipher_kt); +int cipher_kt_mode (const cipher_kt_t *cipher_kt); /** diff --git a/crypto_openssl.c b/crypto_openssl.c index db6b78e..0a41b39 100644 --- a/crypto_openssl.c +++ b/crypto_openssl.c @@ -555,7 +555,7 @@ cipher_kt_block_size (const EVP_CIPHER *cipher_kt) return EVP_CIPHER_block_size (cipher_kt); } -bool +int cipher_kt_mode (const EVP_CIPHER *cipher_kt) { ASSERT(NULL != cipher_kt); diff --git a/crypto_polarssl.c b/crypto_polarssl.c index e7470d5..ac4cadd 100644 --- a/crypto_polarssl.c +++ b/crypto_polarssl.c @@ -313,7 +313,7 @@ cipher_kt_block_size (const cipher_info_t *cipher_kt) return cipher_kt->block_size; } -bool +int cipher_kt_mode (const cipher_info_t *cipher_kt) { ASSERT(NULL != cipher_kt);