Re: Wrong Digest Authentication header when using Windows SSPI
On 25.04.2015 21:30, Daniel Stenberg wrote: On Thu, 23 Apr 2015, LRN wrote: The only hunk that needed to change was the one that adds #include rawstr.h Even without this hunk the code compiles, albeit with warnings (implicit declarations of raw string functions). Even with this hunk i still get *some* warnings: Sorry, I don't mean to be annoying, but since you provided this fine update, could I also bother you to fix the remaining warnings? It'd get the the patch merge ready, and will make the process smoother... I'd be grateful! Here it is. Fixed a typo in the docs (a the - the). Added the prototype to sspi.h. Fits there well enough. Added a cast to (const char*). Patch should apply to 7.42.0 with small offset. Note that i've hacked the patch together manually, it's not from git format-patch. -- O ascii ribbon - stop html email! - www.asciiribbon.org From a69b2a723dcd03cdf74a836b77abc9e9b59bd907 Mon Sep 17 00:00:00 2001 From: Grant Pannell gr...@pannell.net.au Date: Sat, 21 Feb 2015 17:12:11 +1030 Subject: [PATCH] sasl_sspi: Populate the domain from the realm in the challenge message if the user does not specify DOMAIN\User format --- lib/curl_sasl.c | 12 + lib/curl_sasl.h | 11 lib/curl_sasl_sspi.c | 76 +++- 3 files changed, 93 insertions(+), 6 deletions(-) diff --git a/lib/curl_sasl.c b/lib/curl_sasl.c index 4591edb..2ce94d3 100644 --- a/lib/curl_sasl.c +++ b/lib/curl_sasl.c @@ -78,9 +78,6 @@ const struct { #define DIGEST_QOP_VALUE_STRING_AUTH_INT auth-int #define DIGEST_QOP_VALUE_STRING_AUTH_CONF auth-conf -#define DIGEST_MAX_VALUE_LENGTH 256 -#define DIGEST_MAX_CONTENT_LENGTH 1024 - /* The CURL_OUTPUT_DIGEST_CONV macro below is for non-ASCII machines. It converts digest text to ASCII so the MD5 will be correct for what ultimately goes over the network. @@ -92,13 +89,16 @@ const struct { return result; \ } +#endif + +#if !defined(CURL_DISABLE_CRYPTO_AUTH) /* * Returns 0 on success and then the buffers are filled in fine. * * Non-zero means failure to parse. */ -static int sasl_digest_get_pair(const char *str, char *value, char *content, -const char **endptr) +int sasl_digest_get_pair(const char *str, char *value, char *content, + const char **endptr) { int c; bool starts_with_quote = FALSE; @@ -159,7 +159,9 @@ static int sasl_digest_get_pair(const char *str, char *value, char *content, return 0; /* all is fine! */ } +#endif +#if !defined(CURL_DISABLE_CRYPTO_AUTH) !defined(USE_WINDOWS_SSPI) /* Convert md5 chunk to RFC2617 (section 3.1.3) -suitable ascii string*/ static void sasl_digest_md5_to_ascii(unsigned char *source, /* 16 bytes */ unsigned char *dest) /* 33 bytes */ diff --git a/lib/curl_sasl.h b/lib/curl_sasl.h index 0dc7377..88c04ee 100644 --- a/lib/curl_sasl.h +++ b/lib/curl_sasl.h @@ -65,6 +65,11 @@ struct kerberos5data; #define SASL_MECH_STRING_NTLM NTLM #define SASL_MECH_STRING_XOAUTH2XOAUTH2 +#if !defined(CURL_DISABLE_CRYPTO_AUTH) +#define DIGEST_MAX_VALUE_LENGTH 256 +#define DIGEST_MAX_CONTENT_LENGTH 1024 +#endif + enum { CURLDIGESTALGO_MD5, CURLDIGESTALGO_MD5SESS @@ -136,6 +141,12 @@ char *Curl_sasl_build_spn(const char *service, const char *instance); TCHAR *Curl_sasl_build_spn(const char *service, const char *instance); #endif +#if defined(USE_WINDOWS_SSPI) +/* This is used to extract the realm from a challenge message */ +int sasl_digest_get_pair(const char *str, char *value, char *content, +const char **endptr); +#endif + #if defined(HAVE_GSSAPI) char *Curl_sasl_build_gssapi_spn(const char *service, const char *host); #endif diff --git a/lib/curl_sspi.h b/lib/curl_sspi.h --- a/lib/curl_sspi.h +++ b/lib/curl_sspi.h @@ -43,6 +43,10 @@ CURLcode Curl_sspi_global_init(void); void Curl_sspi_global_cleanup(void); +/* This is used to populate the domain in a SSPI identity structure */ +CURLcode Curl_override_sspi_http_realm(const char *chlg, + SEC_WINNT_AUTH_IDENTITY *identity); + /* This is used to generate an SSPI identity structure */ CURLcode Curl_create_sspi_identity(const char *userp, const char *passwdp, SEC_WINNT_AUTH_IDENTITY *identity); diff --git a/lib/curl_sasl_sspi.c b/lib/curl_sasl_sspi.c index 0509b75..698fff9 100644 --- a/lib/curl_sasl_sspi.c +++ b/lib/curl_sasl_sspi.c @@ -40,6 +40,7 @@ #include sendf.h #include strdup.h #include curl_printf.h +#include rawstr.h /* The last #include files should be: */ #include curl_memory.h @@ -276,6 +277,75 @@ CURLcode Curl_sasl_create_digest_md5_message(struct SessionHandle *data, } /* +* Curl_override_sspi_http_realm() +* +* This is used to populate the domain in a SSPI identity structure +* The realm is extracted
Re: Wrong Digest Authentication header when using Windows SSPI
On 23.04.2015 17:27, Daniel Stenberg wrote: On Thu, 23 Apr 2015, LRN wrote: So InitializeSecurityContext() does get the realm reported by the server, but neglects to set it on the request it generates. Does this patch fix it for you? https://github.com/bagder/curl/pull/141 Yes, it does (does not apply to 7.42.0 cleanly though, had to tweak it). -- O ascii ribbon - stop html email! - www.asciiribbon.org 0x922360B0.asc Description: application/pgp-keys signature.asc Description: OpenPGP digital signature --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Wrong Digest Authentication header when using Windows SSPI
The header i get is: Digest username=\testuser\,realm=\\,nonce=\...\,uri=\/\,cnonce=\...\,nc=0001,response=\...\,qop=\auth\,opaque=\...\ (replaced actual hashes with ...). Note the empty realm. The chlg_buf[0].pvBuffer that gets fed to s_pSecFn-InitializeSecurityContext() by Curl_sasl_create_digest_http_message() is: realm=\t...@example.com\,qop=\auth\,nonce=\...\,opaque=\...\ So InitializeSecurityContext() does get the realm reported by the server, but neglects to set it on the request it generates. -- O ascii ribbon - stop html email! - www.asciiribbon.org 0x922360B0.asc Description: application/pgp-keys signature.asc Description: OpenPGP digital signature --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Re: Wrong Digest Authentication header when using Windows SSPI
On 23.04.2015 17:54, Daniel Stenberg wrote: On Thu, 23 Apr 2015, LRN wrote: Yes, it does (does not apply to 7.42.0 cleanly though, had to tweak it). Nice, can you post an updated version of the patch for us to consider to merge? I mostly lack a review and a +1 from someone who knows something about Windows and SSPI, but a works for me goes pretty far too. The attached patch applies to 7.42.0 cleanly. The only hunk that needed to change was the one that adds #include rawstr.h Even without this hunk the code compiles, albeit with warnings (implicit declarations of raw string functions). Even with this hunk i still get *some* warnings: ../../curl-7.42.0/lib/curl_sasl_sspi.c:291:10: warning: no previous prototype for 'Curl_override_sspi_http_realm' [-Wmissing-prototypes] CURLcode Curl_override_sspi_http_realm (const char *chlg, ^ ../../curl-7.42.0/lib/curl_sasl_sspi.c: In function 'Curl_sasl_create_digest_http_message': ../../curl-7.42.0/lib/curl_sasl_sspi.c:448:39: warning: pointer targets in passing argument 1 of 'Curl_override_sspi_http_realm' differ in signedness [-Wpointer-sign] if (Curl_override_sspi_http_realm(digest-input_token, identity)) ^ ../../curl-7.42.0/lib/curl_sasl_sspi.c:291:10: note: expected 'const char *' but argument is of type 'BYTE *' CURLcode Curl_override_sspi_http_realm (const char *chlg, ^ Both should be easy to fix (provide a prototype just before the function to placate gcc; cast the argument as 'const char*') -- O ascii ribbon - stop html email! - www.asciiribbon.org From a69b2a723dcd03cdf74a836b77abc9e9b59bd907 Mon Sep 17 00:00:00 2001 From: Grant Pannell gr...@pannell.net.au Date: Sat, 21 Feb 2015 17:12:11 +1030 Subject: [PATCH] sasl_sspi: Populate the domain from the realm in the challenge message if the user does not specify DOMAIN\User format --- lib/curl_sasl.c | 12 + lib/curl_sasl.h | 11 lib/curl_sasl_sspi.c | 76 +++- 3 files changed, 93 insertions(+), 6 deletions(-) diff --git a/lib/curl_sasl.c b/lib/curl_sasl.c index 4591edb..2ce94d3 100644 --- a/lib/curl_sasl.c +++ b/lib/curl_sasl.c @@ -78,9 +78,6 @@ const struct { #define DIGEST_QOP_VALUE_STRING_AUTH_INT auth-int #define DIGEST_QOP_VALUE_STRING_AUTH_CONF auth-conf -#define DIGEST_MAX_VALUE_LENGTH 256 -#define DIGEST_MAX_CONTENT_LENGTH 1024 - /* The CURL_OUTPUT_DIGEST_CONV macro below is for non-ASCII machines. It converts digest text to ASCII so the MD5 will be correct for what ultimately goes over the network. @@ -92,13 +89,16 @@ const struct { return result; \ } +#endif + +#if !defined(CURL_DISABLE_CRYPTO_AUTH) /* * Returns 0 on success and then the buffers are filled in fine. * * Non-zero means failure to parse. */ -static int sasl_digest_get_pair(const char *str, char *value, char *content, -const char **endptr) +int sasl_digest_get_pair(const char *str, char *value, char *content, + const char **endptr) { int c; bool starts_with_quote = FALSE; @@ -159,7 +159,9 @@ static int sasl_digest_get_pair(const char *str, char *value, char *content, return 0; /* all is fine! */ } +#endif +#if !defined(CURL_DISABLE_CRYPTO_AUTH) !defined(USE_WINDOWS_SSPI) /* Convert md5 chunk to RFC2617 (section 3.1.3) -suitable ascii string*/ static void sasl_digest_md5_to_ascii(unsigned char *source, /* 16 bytes */ unsigned char *dest) /* 33 bytes */ diff --git a/lib/curl_sasl.h b/lib/curl_sasl.h index 0dc7377..88c04ee 100644 --- a/lib/curl_sasl.h +++ b/lib/curl_sasl.h @@ -65,6 +65,11 @@ struct kerberos5data; #define SASL_MECH_STRING_NTLM NTLM #define SASL_MECH_STRING_XOAUTH2XOAUTH2 +#if !defined(CURL_DISABLE_CRYPTO_AUTH) +#define DIGEST_MAX_VALUE_LENGTH 256 +#define DIGEST_MAX_CONTENT_LENGTH 1024 +#endif + enum { CURLDIGESTALGO_MD5, CURLDIGESTALGO_MD5SESS @@ -136,6 +141,12 @@ char *Curl_sasl_build_spn(const char *service, const char *instance); TCHAR *Curl_sasl_build_spn(const char *service, const char *instance); #endif +#if defined(USE_WINDOWS_SSPI) +/* This is used to extract the realm from a challenge message */ +int sasl_digest_get_pair(const char *str, char *value, char *content, +const char **endptr); +#endif + #if defined(HAVE_GSSAPI) char *Curl_sasl_build_gssapi_spn(const char *service, const char *host); #endif diff --git a/lib/curl_sasl_sspi.c b/lib/curl_sasl_sspi.c index 0509b75..698fff9 100644 --- a/lib/curl_sasl_sspi.c +++ b/lib/curl_sasl_sspi.c @@ -40,6 +40,7 @@ #include sendf.h #include strdup.h #include curl_printf.h +#include rawstr.h /* The last #include files should be: */ #include curl_memory.h @@ -276,6 +277,75 @@ CURLcode Curl_sasl_create_digest_md5_message(struct SessionHandle *data
Re: [bagder/curl] efa548: ftp.pm: Made Perl testsuite able to kill Windows p...
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 06.04.2013 14:49, Marc Hoersken wrote: Hi Gisle, 2013/4/6 Gisle Vanem gva...@broadpark.no: ftp.pm: Made Perl testsuite able to kill Windows processes This patch uses a tool from Windows-XP Professional: system(taskkill $filter nul 21); I run Windows-XP Home and have no such tool. AFAICS, one cannot get this from MS w/o having a pro version. Isn't there a better alternative (like pskill)? I am currently testing on Windows 7, so I was not sure if it's available on earlier versions. Thanks for the input. I will have to see if there is a better command to check and kill PIDs. The problem with pskill is that it would have to be manually installed by test users. You can use Win32API Perl module and TerminateProcess or SafeTerminateProcess. Also, CreateJobObject + AssignProcessToJobObject can be used to implement behaviour similar to kill -pid. Obviously, that also requires Win32API module. -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJRYCz6AAoJEOs4Jb6SI2CwOnsH/3UkYOqnawA/21Mr3a0Uar97 x/7E/lmYSFROvDq5l7XWFnMIyjIqXSuo/MrXe1JraCLsWh53JQHjWIyLwNLW+i6z YFQkFnFraT1X7Vw0ej44/28ntID4OgIsBeJ5iyeQz9TG3s09DXTJQSEWTvfQ8dyr ssHQsAArSpW0Rtp8s3sMr0cQx9lIifgqO7+DDcPfWH7OgMFpOD5d0yXc1u0tX5qz lm8kOrN6USPfLaulw7oHZA9S2i8NUCNgWW3Td+H+IZ1fcuu/XUiLVCo/QbOOqm7+ 5qbbuK+XTbXc27gi2LAfZzLrcyiypgpm2vVCTvBOmMZSi91f58ABxI4v5Q0f3LM= =Rpoo -END PGP SIGNATURE- --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Re: [bagder/curl] efa548: ftp.pm: Made Perl testsuite able to kill Windows p...
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 06.04.2013 19:12, Marc Hoersken wrote: Hi LRN, 2013/4/6 LRN lrn1...@gmail.com: You can use Win32API Perl module and TerminateProcess or SafeTerminateProcess. Also, CreateJobObject + AssignProcessToJobObject can be used to implement behaviour similar to kill -pid. Obviously, that also requires Win32API module. I added support for Windows XP by also using tskill to terminate the process. I personally don't like the idea of adding new third-party Perl modules with Windows API dependencies, unless it's a requirement. Any objections? The only advice i can give is to treat tskill failures gracefully. W32 is known to put processes into a state in which they still exist, but cannot be interacted with, probably due to the fact that OS is dismantling them at the moment. Killing such process at that time will fail with Access denied, no matter who you are or what permissions you have. If that happens, i'd advise to wait for a second or two (the best way is to wait on process termination, but that requires Win32API; you can instead just wait for short amounts of time in a loop, then check whether the process is dead; or just wait for the full amount of time, if performance is not a goal), then check if the process is still alive and try again if it still is. -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJRYD70AAoJEOs4Jb6SI2CwTL0H/jnPMHplHvKkV7PAgB3hL7Ir GNDqtvUW8y+vWZBCWd3NQkBvnIwsg2JrkydYcTS7O4X7UaaSvh4pF0NJwQ0fg/Q2 uzfaffrZNPyi6E8PPt6IAuEVXWxP7T112ZB32XQbs+CMncSxjPDqNxLL/HU6M0Bx IebnjSFir/X4dzfKVJDcVKyACiP/ZAn9An1LCizn8axPpOCnjTM2iheIlGHqaUYX XlJiUnBulmOGOjldk4SxUImiF3dSOeZ7uue8xg55nROnGwHIh/9IGuFlL7S9/+qv 4Bl9gOA+wATt+0AjS36gIw/EIlu1xEf1wnUAUApSNUMjdlAERfB0KYakGw5wfhY= =37B/ -END PGP SIGNATURE- --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Re: Test results for W32
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 02.04.2013 18:20, Dan Fandrich wrote: On Tue, Apr 02, 2013 at 05:29:10PM +0400, LRN wrote: 24 285 543 571 1309 fail due to \r\n vs \n difference (my guess is that data is sent over the network with correct EOLS, but curl client prints it in text mode). Is this an issue with how you unpacked the source code? IIRC, the test suite should handle this issue. Someone else can comment on other known issues with running on Windows, but I thought it was working in Cygwin. That's MSYS, not Cygwin. I unpack via tar --lzma -xf, it doesn't mess up file contents. Can't really say what's wrong with other tests. Most of them don't give any output, maybe i haven not compiled appropriate parts of curl that are tested in there? The test suite auto-detects most curl build configurations. Only if you used an exotic configure or compile flag should you need to worry about this (see the REDUCING SIZE section of INSTALL). I configure with --prefix=/mingw \ --enable-debug \ --enable-optimize \ --enable-warnings \ --with-zlib=/mingw \ --without-ssl \ --with-gnutls=/mingw \ --enable-ipv6 \ --build=i686-w64-mingw32 \ --host=i686-w64-mingw32 so that is not the case. The first log showed some signs of improper test suite shutdown in some of the tests. Try making sure all the test suite processes are shut down and running the failing tests one at a time. Is there a way to specify commands that test harness would run in-between tests? I'd use that facility to run ps and kill. Also, use the -d option to look for issues with starting stopping the protocol test servers. Will do -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJRXMKsAAoJEOs4Jb6SI2CwLSUIAKIUyqzjw6qEM6U1Vsf7SRGs sd1mwR3pPfRyX2/30QYSM0R82vfCIwsy2YdIxpgfWT3+spsbCc9n7UJDdavbtIOl VzsVw6bRcPbc9egh2h+8f1W/UE01QCRCgdOmL9AwnHTAtAweD6Nd/D43bShz5FMe RA2rEborU4M0ZDbl/e/GS6OsQ6zKuFb/6RbvTe4PMBZBCAFVajB0571/PCxFwqIL chUx5/XVvjdGGpFuZ9NUpQ6Bmmcb90/pkOMEyB5FJ1anM8tY4IWrBw4pF7xfH7hW pRvehJPk8gB0PmSkXYx/p0hX2epjwlWSNi56DHP0humIhBz60S8zy6I6stFqnck= =2F64 -END PGP SIGNATURE- --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Re: Test results for W32
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 02.04.2013 12:04, Daniel Stenberg wrote: On Tue, 2 Apr 2013, LRN wrote: Here's the output of `make check'. Hopefully, this will be useful. We need a whole lot more details for this to serve any purpose. Start with the first failure and analyze it closer. Then next and next and next... Here's the log with full output. I think i should mention that test 286 seemed to hung up, so i had to kill it. 24 285 543 571 1309 fail due to \r\n vs \n difference (my guess is that data is sent over the network with correct EOLS, but curl client prints it in text mode). Can't really say what's wrong with other tests. Most of them don't give any output, maybe i haven not compiled appropriate parts of curl that are tested in there? -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJRWt0mAAoJEOs4Jb6SI2Cw9QUH/2RgMNA9F2wMtHGdbBcDq2dH 2itR1Zh8yOemSiBsMCj/O5kWF6usUSbNbAXBmrdEde4LuOoIIrrbHc29pTH0+hEF FxB79sic9oxguhbVW20I/nr6YNmIU+ZxYnAw46cCVrFUYeXui3MBcAK8VUIsX/se gd/dI2auPkawdeMJiBpYoWn3RLKXMFfrm8JoN4FBKToZPljHD45cDb5RePobNJXU /4FVNAJUfIHlrVbm1yzEVIQqwlgSue4UPowwKRrFeQXM+z0HTrXH049LMic1DAta EGhsB5+ESKB7CP8i0xXlRkDN+AwOJXGyPTTuiEHuMM78+BrBsW8S7VRZDBYiarM= =DglM -END PGP SIGNATURE- tests.log.xz Description: Binary data --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Test results for W32
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Here's the output of `make check'. Hopefully, this will be useful. -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJRWjIQAAoJEOs4Jb6SI2CwIyMIAKGvYTwSeH35onLevQSqnJtM nLX1PScW97/GChnBQ6kCdY4zF0rT2OHugMYVGWKhKVaHG1m+RGqGIQliE1mSYRj6 wMYwp5ACp4w30PqSkSE7hv/Aw/vWDT1tDeqWo0HzVkfHOMJ5DvV5AsQcDfliFiQb CmX2YS2x4bJoC39dM6Nnt+PUddr4IK5dise9nt7NZUbv3A21TUrNC2tZuxa4B00b BPRGKlEEThPZZqYhhI+aP0gRhJ3eqOM8Ozn3L8GmKj2wuohdMQh/c4Ol32NNjiRL 3hf1ilThB0J1V7ZttyNZ5Mvp7UvsI1ZPSygtFrkxR+CG8UBR9TlOCZ3H+Q+p7UQ= =5IJa -END PGP SIGNATURE- check.log.xz Description: Binary data --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Re: [W32] Testsuite does not build in 7.28.1 anymore
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 21.01.2013 13:55, Daniel Stenberg wrote: On Sat, 19 Jan 2013, LRN wrote: `make check' log is attached. It worked in 7.28.0 (or, at least, 7.27.0). Tried rebuilding everything with --disable-shared, didn't help. In that massive log I could mostly see complaints on various *printf() functions. In which way have they changed between 7.27.0 and 7.28.0? I can't remember any change in that area for a very long time... I don't think the problem is in functions per se. The problem seems to be in the testsuite linking the tests in a way that doesn't quite work on W32. AFAIU, instead of linking to libcurl, as apps would do, test programs are assembled from some of the compiled object files that go into libcurl. So the header tells it to link to dllimport mprintf functions, while the object files provide only normal mprintf functions for internal use. Something like that. I've tested 7.28.0, testsuite doesn't build there too. 7.27.0 testsuite, however, builds correctly. That said, nothing stands out in 7.27.0-7.28.0 diff, and there are no intermediate releases to check. -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJQ/Rn8AAoJEOs4Jb6SI2CwczUH/RMwBZkyGZFXiuGvdU7l9G7N ExKBS/Z7DMhabA+gfjncykhEYtGpqYUIwRRkZrr8kZSVIGf2V/GOui31NdkP4c6u VcLX1hink6VUHf1DbCw6LEBSSABP0OR7ymyY8HOxsivUaKPFUfOE6o7w7OYwoqXR 8j/zKXV5UaK1sOts6G9gxHPxjB5Xk06qfiXyS8Gx2uiYkjRsmnj+Xa02d059rw0A pLRTrmbfXs28R8e3Kh8YlnbbUeozKapLxJyqoxa3dGC4QpiSMnn3tFdjehrRVFGV KEwWN7ffXYdQvEPYRW7OJRdwO3naUhvuwywKGC0VheDqej4LuQRMxu3jw7Tg2v4= =pp0D -END PGP SIGNATURE- --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
Re: [W32] Testsuite does not build in 7.28.1 anymore
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On 22.01.2013 6:31, Guenter wrote: Am 21.01.2013 11:35, schrieb LRN: On 21.01.2013 13:55, Daniel Stenberg wrote: On Sat, 19 Jan 2013, LRN wrote: `make check' log is attached. It worked in 7.28.0 (or, at least, 7.27.0). Tried rebuilding everything with --disable-shared, didn't help. In that massive log I could mostly see complaints on various *printf() functions. In which way have they changed between 7.27.0 and 7.28.0? I can't remember any change in that area for a very long time... I don't think the problem is in functions per se. The problem seems to be in the testsuite linking the tests in a way that doesn't quite work on W32. AFAIU, instead of linking to libcurl, as apps would do, test programs are assembled from some of the compiled object files that go into libcurl. So the header tells it to link to dllimport mprintf functions, while the object files provide only normal mprintf functions for internal use. Something like that. I've tested 7.28.0, testsuite doesn't build there too. 7.27.0 testsuite, however, builds correctly. That said, nothing stands out in 7.27.0-7.28.0 diff, and there are no intermediate releases to check. you should try a build with --disable-shared Been there, done that. -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJQ/j0YAAoJEOs4Jb6SI2Cw2QwIAK4QXjCW6gS77ubpZVby00yv NXK5xmbshV71fzw7ZSL3htuc0zySE+bUyPC4J+kYrz6XU5+1iN7oEcAO83YVBuqj P231A0/MG7cLmfolJfj7PwNl7Eqc8xMqmrWSdmYjxosr540ELtVvFADsyJyhe/P3 uQ3OT7G3mRXVvX7Vps2LebwotDMtHyRBRq88JFYAnxFYe5ef4mekwA8WB5nq2nK/ 0d00QsmAp95jKJtk91FGISbIxUlFCiAVpDh4U8AShRZ/6KugZsK0mimWPnI8A8KG P+wD86JhlrN9l/lkg3tj2Zr8Hw/yhWt7fMJWPHuYAj3qJk1UpnzP3JPX8uRHbvg= =Ex3D -END PGP SIGNATURE- --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html
[W32] Testsuite does not build in 7.28.1 anymore
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 `make check' log is attached. It worked in 7.28.0 (or, at least, 7.27.0). Tried rebuilding everything with --disable-shared, didn't help. -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (MingW32) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQEcBAEBAgAGBQJQ+jTvAAoJEOs4Jb6SI2CwKbEH/16oLLXPGpgeCcJiHOXgnAqK yR4numMCwDX9fegKV9hjD5AcI1Jw8VOH7OUuEZV6NMoHx26PPkXUHGCAQyi6HyDi Si7MfEzrholUtj2C0w3MZ0B5EA0W4ey/NPZMayPFlCPoKgH8M8+qw30I7mKw3jUA HgumQF1X5FVt3ryTrhDxWlqwVcNbVQJcaVfNyI9Q1d2OGE+z/ZyVlS1n1rJ7PfVh 4aqq7pz61DbwV4RiBQ7ETM26voUYVTrBYsxNU7G6eb4Vkp43QYUQc3dT3tMKovza gbf8NcdPvkNkowvP5I+AE/TIOckUhSJVLdSg0JXJZ07fLN4JWHYoaQIGh7DxuJU= =KGez -END PGP SIGNATURE- checklog.xz Description: Binary data --- List admin: http://cool.haxx.se/list/listinfo/curl-library Etiquette: http://curl.haxx.se/mail/etiquette.html