The test for libgmp in the configure script for res-rsa doesn't work. In new versions of gmp, the gmp.h header files #defines functions to different names, like mpz_init() to __gmpz_init(). It looks like the autoconf test for libgmp doesn't include gmp.h and so continues to look for mpz_init in libgmp. I don't know autoconf, but I found a dated fix on the web: http://www.kaffe.org/pipermail/kaffe/2000-November/090303.html I adapted this to res-rsa and it appears to work. I'm attaching a darcs patch to fix this in configure.in. I'm not include the change to configure, since it is a generated file. -------------- next part -------------- New patches: [resrsa-gmp-autoconf.darcs Trent Piepho <xyzzy at speakeasy.org>**20060522210616 The test for gmp in the res-rsa autoconf file failed for gmp version 3. In gmp3, functions are #define'd to different names in gmp.h, for example mpz_init() becomes __gmpz_init(). Since the autoconf test doesn't include gmp.h, it didn't find the new names. The fix, copied from http://www.kaffe.org/pipermail/kaffe/2000-November/090303.html, is to just do a new test with the gmp3 names if the first test fails. ] < > { hunk ./Vanilla/ChangeLog 1 +Mon May 22 14:03:38 2006 Trent Piepho <xyzzy at speakeasy.org> + * res-rsa/configure.in The check for gmp fails for gmp version 3, + fix it by adding a secondary check for the gmp3 symbol names. + Mon May 22 10:53:03 2006 Jimmy Huang <jimmyhua73 at yahoo.com> * robotd/assault.c added extra check not to bomb your own armies. hunk ./Vanilla/res-rsa/configure.in 112 LDOLD="$LDFLAGS" LDFLAGS="$GMP_LIB -lgmp" AC_CHECK_LIB(gmp, mpz_fdiv_q_ui, AC_DEFINE(HAVE_LIB_GMP2) GMP_VER=2) + if test $ac_cv_lib_gmp_mpz_fdiv_q_ui = no; then + dnl with gmp3 it's a #define, use real function name + AC_CHECK_LIB(gmp, __gmpz_fdiv_q_ui, AC_DEFINE(HAVE_LIB_GMP2) GMP_VER=2) + fi LDFLAGS="$LDOLD" fi hunk ./Vanilla/res-rsa/configure.in 130 if test $GMP_VER = 1; then AC_CHECK_LIB(gmp, mpz_init, AC_DEFINE(HAVE_LIB_GMP), GMP_VER=0) + if test $ac_cv_lib_gmp_mpz_init = no; then + dnl with gmp3 it's a #define, use real function name + AC_CHECK_LIB(gmp, __gmpz_init, AC_DEFINE(HAVE_LIB_GMP2) GMP_VER=2) + fi fi # Check for GMP2 hunk ./Vanilla/res-rsa/configure.in 139 if test $GMP_VER = 1; then AC_CHECK_LIB(gmp, mpz_fdiv_q_ui, AC_DEFINE(HAVE_LIB_GMP2) GMP_VER=2) + if test $ac_cv_lib_gmp_mpz_fdiv_q_ui = no; then + dnl with gmp3 it's a #define, use real function name + AC_CHECK_LIB(gmp, __gmpz_fdiv_q_ui, AC_DEFINE(HAVE_LIB_GMP2) GMP_VER=2) + fi fi fi } Context: [resolve conflicts with changelog Stephen Thorne <stephen at thorne.id.au>**20060522011533] [fix two second delay on client connection if daemon not running changelog quozl at us.netrek.org**20060521123415] [add game pause/resume/terminate tool, setgame changelog quozl at us.netrek.org**20060521115308] [add max duplicate ip count for pickup play changelog quozl at us.netrek.org**20060519080824] [add install-ntserv target for live updates changelog quozl at us.netrek.org**20060519061157] [mute banned observers changelog quozl at us.netrek.org**20060519042833] [transcription of ChangeLog data for Jimmy's patch Stephen Thorne <stephen at thorne.id.au>**20060522005345] [robotd-fixes1.dpatch jimmyhua73 at yahoo.com**20060521152116 changelog entry here. * robotd/assault.c added extra check not to bomb your own armies. Bots will try to bomb own armies when 2 carriers drop on same planet. And planet ends up with >4 armies. * robotd/decide.c changed some tabbing * robotd/decide.c corrected logic in check_take function. Robotd now takes neutral planets. Before it had a tendency to ignore neutral planets. Also used to crash out when no more enemy planets. Both are fixed with this. * robotd/main.c, shmem.c, socket.c, update_players.c removed mfprintf function. replaced with fprintf function until such time that mfprintf function is fixed so it doesn't crash. Bandaid. * robotd/update_players.c changed some tabbing, added some missing braces. * robotd/update_players.c changed closest_planet() function to return the current closest planet instead of NULL when it cannot find a closer planet. * robotd/util.c added comments to reflect that mfprintf causes crashes for as yet unknown reasons. ] [fix two second delay on client connection if daemon not running quozl at us.netrek.org**20060521123246 Remove the two second delay experienced by clients that connect to a server on which the daemon is not running. Also fix a cause of initial connection failing, which happens if the daemon does not start up within the hoped-for interval. Solution is implemented using the SIGUSR1 signal, but only during the initialisation window between when ntserv forks the daemon and the daemon completes initialisation. The signal is not used at other times. * ntserv/daemonII.c (main): send a SIGUSR1 to parent process (ntserv) once initialisation of shared memory is completed and regular SIGALRMs are about to commence. * ntserv/openmem.c (openmem): remove two second delay that was used to hope for daemon to initialise. * ntserv/openmem.c (startdaemon): add a wait for SIGUSR1. ] [add game pause/resume/terminate tool, setgame quozl at us.netrek.org**20060521115034 * tools/setgame.c: add script utility for pausing, resuming, and terminating the game. * tools/Makefile.in: add setgame target. * ntserv/input.c (input): send bad version packet to terminate client as soon as a GU_GAMEOK termination is requested. * ntserv/daemonII.c (udplayers): terminate daemon as soon as possible if a GU_GAMEOK termination is requested. Change nplayers to nfree, to better explain what the variable is. * include/struct.h (game_ok): remove unused macro. ] [add max duplicate ip count for pickup play quozl at us.netrek.org**20060519075351 Adds a new etc/sysdef option DUPLICATES which is the maximum number of pickup player slots allowed from the same IP address before the next connection is given the Q32 response. This is a denial of service filtering feature. The default is 3. Set this above 16 to make it ineffective. Set this to less than half of the number of players per side, otherwise abusers may not be ejected or banned by majority vote because a majority would be impossible. Does not affect robots, unless they join via the pickup queue ... which would be a misconfiguration. Robots normally join via a special queue. ] [add install-ntserv target for live updates quozl at us.netrek.org**20060519060623 Addition of a target install-ntserv to the ntserv Makefile so that a server operator can do a live update of ntserv during a running game. The target moves the existing in-use ntserv binary into an archive tree before installing the new ntserv binary. This makes it possible to manually back out of an installation if the new binary breaks badly. Other methods to test a new ntserv binary include running it manually for connection back to the client. See ntserv command line options. ] [mute banned observers quozl at us.netrek.org**20060519042723 Banned observers are allowed in and can speak, which was a surprise, but the code has always allowed this since observers were added. Change to prevent banned observers from talking. Simplify ban logic further. ] [fix changelog conflict from att patch Stephen Thorne <stephen at thorne.id.au>**20060519024319] [Transwarp war errormsg williamb at its.caltech.edu**20060519021844] [ATT fixes williamb at its.caltech.edu**20060518081658] [patch acceptance policy regarding patch name and comment quozl at us.netrek.org**20060519015208 Define a new policy of not changing ChangeLog and NEWS until release. ] [resolve ChangeLog conflict again Stephen Thorne <stephen at thorne.id.au>**20060518095505] [add setteam utility source quozl at us.netrek.org**20060518061257 The actual source now, so that builds complete. ] [add setteam utility quozl at us.netrek.org**20060518055920 Add a utility for viewing and changing the starbase reconstruction timer and surrender timer, both of which are in struct teams. Although these timers can be manipulated by other means, setteam makes it easier to script. Some common code now exists in setplanet.c and setteam.c that deserves factorisation. ] [resolve ChangeLog conflict - this is getting annoying Stephen Thorne <stephen at thorne.id.au>**20060518095316] [adding aclocal.m4 Stephen Thorne <stephen at thorne.id.au>**20060518020621] [resolve conflict Stephen Thorne <stephen at thorne.id.au>**20060518034143] [move wdt reset from SIGALRM handler to input handler quozl at us.netrek.org**20060518024416 Slots may jam in POUTFIT after KWINNER if the user terminates the client ungracefully ... such that the TCP connection is gone but the ntserv lives on. During this time, daemon synchronisation still sends SIGALRM to the ntserv, and ntserv's SIGALRM handler still resets the p_ghostbust timer. Prototype solution is to move p_ghostbust watchdog timer reset from within the SIGALRM handler to the section of main loop that is executed when input arrives from the client. Monitoring the p_ghostbust timer with gdb shows it is now reset for a dormant client on receipt of ping responses. ] [improve the autoconf system to use AC_PROG_LIBTOOL and @RANLIB@ correctly. Stephen Thorne <stephen at thorne.id.au>**20060518014309 * tests/build: Set the configure and mktrekon files executable, just in case. * configure.in: Add AC_PROG_LIBTOOL macro. * ntserv/Makefile.in: Use the @RANLIB@ and @RANLIB_FLAGS@ macros. ] [fix the OSX build by passing the -c flag to ranlib. Stephen Thorne <stephen at thorne.id.au>**20060517111623] [Fix style of some code patched by Jimmy, as per Trent's code review Stephen Thorne <stephen at thorne.id.au>**20060517030403] [updateplayers.dpatch jimmyhua73 at yahoo.com**20060516075315 Treats humans and bots differently. Humans with a high planet rating > 4 are assumed to carry once they have a kill. Humans alive longer than 5 minutes with 1 kill are also assumed to carry. If you have 2 kills, the time they assume you carry drops to 2.5 minutes... Uses the OggV packet and "robot!" login to ID the robots. Needed to change server source code and newbie.c source to help support this. All those changes are already there if you have the newbiebetter.dpatch. Jimmy ] [newbie better 2 changelog fix quozl at us.netrek.org**20060516070248 Applying the principles of the Software Release Practice HOWTO http://www.tldp.org/HOWTO/Software-Release-Practice-HOWTO/ ] [newbiebetter.dpatch jimmyhua73 at yahoo.com**20060516071923 newbie now will generate t-mode games randomly any team against any other. Also, won't start diagonal games. newbie starts up bots with -g option so server knows about the bots. genspkt.c also modified so server tells clients which clients are newbie bots. Merlin still nukes bots based on "robot!" login. So, if you want to pretend to be a bot, you may get nuked. Robots will ID each other based both on "robot!" login and also OggV packet. If you are the only player on a server and have a "robot!" login... The newbie game ends... ] [changelog for Compilation fix for gcc4.0 quozl at us.netrek.org**20060516040416] [Compilation fix for gcc4.0 williamb at its.caltech.edu**20060516015945 Fix for jimmy's code with gcc4.0, compiler doesn't like variables declared in this way, so moved the int declare to top of the function. This patch contains the following changes: M ./Vanilla/robotd/assault.c -1 +1 ] [enable CONTINUOUS_MOUSE by default quozl at us.netrek.org**20060515223159 Upgrade of Continuum to 2.11.0 revealed a regression caused by not adopting the features file from the previous server. ] [post-release notes for 2.11.0 quozl at us.netrek.org**20060515100328 Changed the notes regarding the release process, to assist self or others for next release. ] [TAG 2.11.0 quozl at us.netrek.org**20060515091704] Patch bundle hash: 7b84a2950c1b3d2cb577473ecc19555f608007de