Re: jk 1.2.6 to be tagged - showstopper
Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ I've just tried to build on Linux, and it breaks in jki_connect.c on two boxes, SuSE 7.2 and SuSE9, on the first because no in_addr_t, on the second because its defined as struct and not ulong; furthermore I had exactly the same issue with another project, and we killed in_addr_t usage. If you do a Google for in_addr_t you can directly see that this is a problem on many platforms, seems to me that some have it defined as unsigned long, others as struct, and the rest doesnt have it defined at all... I will this weekend examine this further, unless someone beats me... but I think this is a big showstopper, and we should probably hold the release until this is fixed... Well I didn't have this problem on the 3 differents Liux distros : - Fedora Core 2 - Suse 9.0 (i386) - Suse 8.0 (PPC) - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ JFC's portable.h breaks NetWare too; should I now extend the #ifndef like Henry did? I don't see what is wrong: jk/native/common/portable.h should be created from portable.h.in when doing configure. I guess there are probably more platforms which dont have portable.h, so it's perhaps a better approach to do: #ifdef HAVE_PORTABLE_H #include portable.h #endif otherwise it would look like this: #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #else #if defined(NETWARE) !defined(__NOVELL_LIBC__) #include whatever is necessary for getting the vsnprintf, snprint replacements #else #define HAVE_VSNPRINTF #define HAVE_SNPRINTF #endif personally I think its not needed anyway: if we have a good implementation for vsnprintf and friends, then why we do not trust it and use it in general over all platforms?? What's the benefit of using the OS APIs if present? Guenter. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
Henri Gomez wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ I've just tried to build on Linux, and it breaks in jki_connect.c on two boxes, SuSE 7.2 and SuSE9, on the first because no in_addr_t, on the second because its defined as struct and not ulong; furthermore I had exactly the same issue with another project, and we killed in_addr_t usage. If you do a Google for in_addr_t you can directly see that this is a problem on many platforms, seems to me that some have it defined as unsigned long, others as struct, and the rest doesnt have it defined at all... I will this weekend examine this further, unless someone beats me... but I think this is a big showstopper, and we should probably hold the release until this is fixed... Well I didn't have this problem on the 3 differents Liux distros : - Fedora Core 2 - Suse 9.0 (i386) - Suse 8.0 (PPC) I see another problem autom4te.cache should be removed before doing the tarball. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
jean-frederic clere wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ JFC's portable.h breaks NetWare too; should I now extend the #ifndef like Henry did? I don't see what is wrong: jk/native/common/portable.h should be created from portable.h.in when doing configure. Well iSeries and Windows didn't use autoconf - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: jk 1.2.6 to be tagged
jean-frederic clere wrote: JFC's portable.h breaks NetWare too; should I now extend the #ifndef like Henry did? That's the quickest way. The proper would be to have protable.hw and portable.nw (like APR, Apache, etc... have). I don't see what is wrong: jk/native/common/portable.h should be created from portable.h.in when doing configure. We don't use configure. MT. smime.p7s Description: S/MIME cryptographic signature
Re: jk 1.2.6 to be tagged - showstopper
jean-frederic clere wrote: Henri Gomez wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ I've just tried to build on Linux, and it breaks in jki_connect.c on two boxes, SuSE 7.2 and SuSE9, on the first because no in_addr_t, on the second because its defined as struct and not ulong; furthermore I had exactly the same issue with another project, and we killed in_addr_t usage. If you do a Google for in_addr_t you can directly see that this is a problem on many platforms, seems to me that some have it defined as unsigned long, others as struct, and the rest doesnt have it defined at all... I will this weekend examine this further, unless someone beats me... but I think this is a big showstopper, and we should probably hold the release until this is fixed... Well I didn't have this problem on the 3 differents Liux distros : - Fedora Core 2 - Suse 9.0 (i386) - Suse 8.0 (PPC) I see another problem autom4te.cache should be removed before doing the tarball. I run the buildconf.sh. May be that's the problem ? - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ JFC's portable.h breaks NetWare too; should I now extend the #ifndef like Henry did? I guess there are probably more platforms which dont have portable.h, so it's perhaps a better approach to do: #ifdef HAVE_PORTABLE_H #include portable.h #endif otherwise it would look like this: #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #else #if defined(NETWARE) !defined(__NOVELL_LIBC__) #include whatever is necessary for getting the vsnprintf, snprint replacements #else #define HAVE_VSNPRINTF #define HAVE_SNPRINTF #endif personally I think its not needed anyway: if we have a good implementation for vsnprintf and friends, then why we do not trust it and use it in general over all platforms?? What's the benefit of using the OS APIs if present? There is also reports for Solaris 8.0 where portable.h is not present ;( - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
Henri Gomez wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ JFC's portable.h breaks NetWare too; should I now extend the #ifndef like Henry did? I guess there are probably more platforms which dont have portable.h, so it's perhaps a better approach to do: #ifdef HAVE_PORTABLE_H #include portable.h #endif otherwise it would look like this: #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #else #if defined(NETWARE) !defined(__NOVELL_LIBC__) #include whatever is necessary for getting the vsnprintf, snprint replacements #else #define HAVE_VSNPRINTF #define HAVE_SNPRINTF #endif personally I think its not needed anyway: if we have a good implementation for vsnprintf and friends, then why we do not trust it and use it in general over all platforms?? What's the benefit of using the OS APIs if present? There is also reports for Solaris 8.0 where portable.h is not present ;( May be we should have a default ./common/portable.h containing: /* On most platform this file is overwritten when doing configure */ #define PACKAGE mod_jk #define VERSION 1.2.6 BTW: The version is wrong in configure.in should be VERSION=1.2.6 - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
Henri Gomez wrote: jean-frederic clere wrote: Henri Gomez wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ I've just tried to build on Linux, and it breaks in jki_connect.c on two boxes, SuSE 7.2 and SuSE9, on the first because no in_addr_t, on the second because its defined as struct and not ulong; furthermore I had exactly the same issue with another project, and we killed in_addr_t usage. If you do a Google for in_addr_t you can directly see that this is a problem on many platforms, seems to me that some have it defined as unsigned long, others as struct, and the rest doesnt have it defined at all... I will this weekend examine this further, unless someone beats me... but I think this is a big showstopper, and we should probably hold the release until this is fixed... Well I didn't have this problem on the 3 differents Liux distros : - Fedora Core 2 - Suse 9.0 (i386) - Suse 8.0 (PPC) I see another problem autom4te.cache should be removed before doing the tarball. I run the buildconf.sh. May be that's the problem ? The problem is in buildconf.sh: I have fixed it. If you leave autom4te.cache in the distribution and someone makes automake using a different automake than the one you have used when preparing the tarball *very* strange things will happend... - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
jean-frederic clere wrote: Henri Gomez wrote: jean-frederic clere wrote: Henri Gomez wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ I've just tried to build on Linux, and it breaks in jki_connect.c on two boxes, SuSE 7.2 and SuSE9, on the first because no in_addr_t, on the second because its defined as struct and not ulong; furthermore I had exactly the same issue with another project, and we killed in_addr_t usage. If you do a Google for in_addr_t you can directly see that this is a problem on many platforms, seems to me that some have it defined as unsigned long, others as struct, and the rest doesnt have it defined at all... I will this weekend examine this further, unless someone beats me... but I think this is a big showstopper, and we should probably hold the release until this is fixed... Well I didn't have this problem on the 3 differents Liux distros : - Fedora Core 2 - Suse 9.0 (i386) - Suse 8.0 (PPC) I see another problem autom4te.cache should be removed before doing the tarball. I run the buildconf.sh. May be that's the problem ? The problem is in buildconf.sh: I have fixed it. If you leave autom4te.cache in the distribution and someone makes automake using a different automake than the one you have used when preparing the tarball *very* strange things will happend... And for the portable.h was could we do ? - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
jean-frederic clere wrote: Henri Gomez wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ JFC's portable.h breaks NetWare too; should I now extend the #ifndef like Henry did? I guess there are probably more platforms which dont have portable.h, so it's perhaps a better approach to do: #ifdef HAVE_PORTABLE_H #include portable.h #endif otherwise it would look like this: #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #else #if defined(NETWARE) !defined(__NOVELL_LIBC__) #include whatever is necessary for getting the vsnprintf, snprint replacements #else #define HAVE_VSNPRINTF #define HAVE_SNPRINTF #endif personally I think its not needed anyway: if we have a good implementation for vsnprintf and friends, then why we do not trust it and use it in general over all platforms?? What's the benefit of using the OS APIs if present? There is also reports for Solaris 8.0 where portable.h is not present ;( May be we should have a default ./common/portable.h containing: /* On most platform this file is overwritten when doing configure */ #define PACKAGE mod_jk #define VERSION 1.2.6 BTW: The version is wrong in configure.in should be VERSION=1.2.6 ok, I'll reworked on CVS to fix all of these and making now new tarballs. The jk 1.2.6 release (with fixes) was tagged as JK_1_2_6_rel Regards - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
Henri Gomez wrote: jean-frederic clere wrote: Henri Gomez wrote: jean-frederic clere wrote: Henri Gomez wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ I've just tried to build on Linux, and it breaks in jki_connect.c on two boxes, SuSE 7.2 and SuSE9, on the first because no in_addr_t, on the second because its defined as struct and not ulong; furthermore I had exactly the same issue with another project, and we killed in_addr_t usage. If you do a Google for in_addr_t you can directly see that this is a problem on many platforms, seems to me that some have it defined as unsigned long, others as struct, and the rest doesnt have it defined at all... I will this weekend examine this further, unless someone beats me... but I think this is a big showstopper, and we should probably hold the release until this is fixed... Well I didn't have this problem on the 3 differents Liux distros : - Fedora Core 2 - Suse 9.0 (i386) - Suse 8.0 (PPC) I see another problem autom4te.cache should be removed before doing the tarball. I run the buildconf.sh. May be that's the problem ? The problem is in buildconf.sh: I have fixed it. If you leave autom4te.cache in the distribution and someone makes automake using a different automake than the one you have used when preparing the tarball *very* strange things will happend... And for the portable.h was could we do ? Adding portable.hnw for netware and portable.hw for windoze and #ifdef for AS400 is the clean way. The quick way is adding: +++ #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #endif +++ - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ I've just tried to build on Linux, and it breaks in jki_connect.c on two boxes, SuSE 7.2 and SuSE9, on the first because no in_addr_t, on the second because its defined as struct and not ulong; furthermore I had exactly the same issue with another project, and we killed in_addr_t usage. If you do a Google for in_addr_t you can directly see that this is a problem on many platforms, seems to me that some have it defined as unsigned long, others as struct, and the rest doesnt have it defined at all... I will this weekend examine this further, unless someone beats me... I don't get the problems on my 7.2 and 9.0 ;-( but I think this is a big showstopper, and we should probably hold the release until this is fixed... Guenter. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
jean-frederic clere wrote: Henri Gomez wrote: jean-frederic clere wrote: Henri Gomez wrote: jean-frederic clere wrote: Henri Gomez wrote: Günter Knauf wrote: Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ I've just tried to build on Linux, and it breaks in jki_connect.c on two boxes, SuSE 7.2 and SuSE9, on the first because no in_addr_t, on the second because its defined as struct and not ulong; furthermore I had exactly the same issue with another project, and we killed in_addr_t usage. If you do a Google for in_addr_t you can directly see that this is a problem on many platforms, seems to me that some have it defined as unsigned long, others as struct, and the rest doesnt have it defined at all... I will this weekend examine this further, unless someone beats me... but I think this is a big showstopper, and we should probably hold the release until this is fixed... Well I didn't have this problem on the 3 differents Liux distros : - Fedora Core 2 - Suse 9.0 (i386) - Suse 8.0 (PPC) I see another problem autom4te.cache should be removed before doing the tarball. I run the buildconf.sh. May be that's the problem ? The problem is in buildconf.sh: I have fixed it. If you leave autom4te.cache in the distribution and someone makes automake using a different automake than the one you have used when preparing the tarball *very* strange things will happend... And for the portable.h was could we do ? Adding portable.hnw for netware and portable.hw for windoze and #ifdef for AS400 is the clean way. The quick way is adding: +++ #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #endif +++ You understand why I wish a new JK using ONLY APR. We spend too many times in such dreaded issues ;( - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
Hi, Adding portable.hnw for netware and portable.hw for windoze and #ifdef for AS400 is the clean way. The quick way is adding: +++ #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #endif +++ unfortunately it isnt that simple cause I have _two_ different NetWare architectures, a modern one with vsnprintf friends, and the older one without where I need the replacements: for NetWare netscape, jni, apache-1.3 need replacement; only apache-2.0 has vsnprintf coming from the OS... I had already fixed that in jk_global.h, but if we stay with portable.h I would like to create protable entirely from the Makefile rather than having a protable.hnw for the above reasons - otherwise I would need _two_ : a protable.hnw1 and portable.hnw2; so better I generate it I'm very short in time today, but see what I can add to the Makefile Guenter. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
Günter Knauf wrote: Hi, Adding portable.hnw for netware and portable.hw for windoze and #ifdef for AS400 is the clean way. The quick way is adding: +++ #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #endif +++ unfortunately it isnt that simple cause I have _two_ different NetWare architectures, a modern one with vsnprintf friends, and the older one without where I need the replacements: for NetWare netscape, jni, apache-1.3 need replacement; only apache-2.0 has vsnprintf coming from the OS... I had already fixed that in jk_global.h, but if we stay with portable.h I would like to create protable entirely from the Makefile rather than having a protable.hnw for the above reasons - otherwise I would need _two_ : a protable.hnw1 and portable.hnw2; so better I generate it I'm very short in time today, but see what I can add to the Makefile Well it will be in jk 1.2.7, if we ever release it since I'm a little borred to see that we still battling against problems fixed by the use of APR. I don't know yet what will do with mod_proxy/ajp_proxy, since it appears that relay via mod_proxy is today 50% slower than with JK and we have to wait Apache 2.1 to see it, but I feel we should seriously study our initial idea of mod_ajp using APR and Apache 2.x as main target. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
Hi, I don't get the problems on my 7.2 and 9.0 ;-( hmm, maybe it's a 7.1, or 7.0 box; I will check; but I looked already with man inet_addr and found ulong as return value, and not in_add_t; do we have any chance that configure detects if we have in_addr_t or not? Guenter. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged - showstopper
Gnter Knauf wrote: Hi, I don't get the problems on my 7.2 and 9.0 ;-( hmm, maybe it's a 7.1, or 7.0 box; I will check; but I looked already with man inet_addr and found ulong as return value, and not in_add_t; do we have any chance that configure detects if we have in_addr_t or not? Yes, I will look how it is done in APR. Guenter. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
Henri Gomez wrote: I'll tag jk 1.2.6 by 17h CET. So if you have any showstopper, send me an email :))) JTC tagged as JK_1_2_6 Now back to jk 1.2.7-dev tarball to be released soon. Thanks to various commiters to provide the usual binaries... Regards - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
Henri Gomez wrote: Henri Gomez wrote: I'll tag jk 1.2.6 by 17h CET. So if you have any showstopper, send me an email :))) JTC tagged as JK_1_2_6 Now back to jk 1.2.7-dev tarball to be released soon. Source tarballs available at : http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ Thanks to check and make some binaries. PS: Mladen I'm more than interested in the Windows binaries for Apache 2.0.50 - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ JFC's portable.h breaks NetWare too; should I now extend the #ifndef like Henry did? I guess there are probably more platforms which dont have portable.h, so it's perhaps a better approach to do: #ifdef HAVE_PORTABLE_H #include portable.h #endif otherwise it would look like this: #if !defined(WIN32) !defined(AS400) !defined(NETWARE) #include portable.h #else #if defined(NETWARE) !defined(__NOVELL_LIBC__) #include whatever is necessary for getting the vsnprintf, snprint replacements #else #define HAVE_VSNPRINTF #define HAVE_SNPRINTF #endif personally I think its not needed anyway: if we have a good implementation for vsnprintf and friends, then why we do not trust it and use it in general over all platforms?? What's the benefit of using the OS APIs if present? Guenter. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: jk 1.2.6 to be tagged
Hi, http://jakarta.apache.org/~hgomez/jakarta-tomcat-connectors-jk-1.2.6/ NetWare binaries for Apache 1.3.x, 2.0.x, 2.1-dev, Netscape are available for testing: http://www.gknw.com/test/jtc/ Guenter. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]