[BACK]Return to ChangeLog CVS log [TXT] [DIR] Up to [local] / rpl

Diff for /rpl/ChangeLog between versions 1.51 and 1.67

version 1.51, 2010/09/23 08:56:56 version 1.67, 2011/06/22 12:48:29
Line 1 Line 1
 * Wed Sep 15 2010 BERTRAND Joël <rpl@systella.fr> rpl-4.0.19 (RPL/2)  * Tue Jun 21 2011 BERTRAND Joël <rpl@systella.fr> rpl-4.1.0.prerelease.2 (RPL/2)
   - memory caching is added to new variable tree.
   * Tue Jun 21 2011 BERTRAND Joël <rpl@systella.fr> rpl-4.1.0.prerelease.1 (RPL/2)
   - SPAWN is now usable with new variable management;
   - MEM is rewritten;
   - PRUSR is rewritten;
   - merge 4_0 branch into HEAD;
   - CLUSR is rewritten;
   - -p flag is now usable;
   - New benchmark:
   cauchy:[~/rpl/bench] > ./bench.rpl 
   +++RPL/2 (R) version 4.1.0.prerelease.1 (lundi 20/06/2011, 23:00:45 CEST)
   +++Copyright (C) 1989 à 2010, 2011 BERTRAND Joël
   Number of scalar operations per second and per core : 2625369.22710792
   * Mon Jun 20 2011 BERTRAND Joël <rpl@systella.fr> rpl-4.1.0.prerelease.0 (RPL/2)
   - New variable management is written. Please note that -p flag does not work
   and all multithread/multitask capabilities have been disabled until this
   new subsystem shall be validated.
   * Thu May 09 2011 BERTRAND Joël <rpl@systella.fr> rpl-4.0.23 (RPL/2)
   - major bug is fixed in STD format (1E7 sq prints 1);
   - major bug is fixed in READ intrinsic (socket). This bug was introduced
   by a new feature in glibc.
   * Thu Apr 14 2011 BERTRAND Joël <rpl@systella.fr> rpl-4.0.22 (RPL/2)
   - by default, RPL/2 uses "NONE" IMPLICIT when it's started in batch mode and
   "ALL" IMPLICIT when it runs in interactive mode;
   - when "NONE" IMPLICIT is set, STO intrinsic raises an error when it is used
   to create or modify a global variable. Now, with "NONE" IMPLICIT, you have
   to use SAVE to manipulate global variables;
   - replace erroneous TAB by TBL in evaluation() and formateur_fichier()
   subroutines;
   - fix major memory access violation in evaluation subroutine. Only occurs
   when 'X(i,j)' is evaluated on a matrix and when (i,j) points out of bounds;
   - fix major bug in evaluation subroutine. 'X(i)' returns execution error
   when X was a vector or a table.
   * Sun Apr 04 2011 BERTRAND Joël <rpl@systella.fr> rpl-4.0.21 (RPL/2)
   - Gnuplot is upgraded;
   - flag 34 is now used to enable or disable control character evaluation;
   - major bug is fixed in P->R intrinsic function when argument type is
   complex and degree mode is set;
   - L->T and T->L are added;
   - OLD filetype is fixed;
   - FLOW socket support is added;
   - FLOW sequential files are added;
   - new escape sequence is added to handle flow files (\x + hexadecimal code).
   Known escape sequences are \", \b, \n, \t, \x?? and \\.
   * Sun Mar 06 2011 BERTRAND Joël <rpl@systella.fr> rpl-4.0.20 (RPL/2)
   - Ncurses is upgraded to 5.8;
   - ALARM is fixed by filling isdst field;
   - catch ECONNRESET in write intrinsic to return an execution error instead of
   system one;
   - fix a typo in HELP intrinsic;
   - Openssl is upgraded to 1.0.0d, sqlite to 3.7.5 and readline to 6.2;
   - Lapack is upgraded to 3.3.0;
   - libmtmalloc is now default Solaris allocator;
   - ncurses is upgraded;
   - regression introduced by 4.0.19 is fixed in signal management.
   Interruption1() could be called from itself and raises bus error (on sparc) or
   segmentation fault;
   - sqlite is upgraded to 3.7.4;
   - OpenSSL is upgraded to 1.0.0c;
   - syntax error is correctly checked ( (2+,0) does not crash RPL/2 anymore);
   - characters are tested in recherche_type() to avoid names like '12+';
   - gnuplot is upgraded to 4.4.2 due to major bugs.
   * Thu Sep 23 2010 BERTRAND Joël <rpl@systella.fr> rpl-4.0.19 (RPL/2)
   - due to a very strange bug I've only seen on Solaris, ptmalloc3 is removed.
   Solaris libresolv randomly uses ptmalloc3 and system allocator and makes heap
   corruption. As internal memory management have been improved, a best fit
   allocator like ptmalloc3 is not required anymore;
 - gnuplot is upgraded to 4.4.1;  - gnuplot is upgraded to 4.4.1;
 - deadlock is fixed in signal handlers: same mutex was used in standard  - deadlock is fixed in signal handlers: same mutex was used in standard
 subroutines (without protected sections by sigprocmask() or pthread_sigmask())  subroutines (without protected sections by sigprocmask() or pthread_sigmask())
Removed from v.1.51  
changed lines
  Added in v.1.67

CVSweb interface <joel.bertrand@systella.fr>