rl-glue - Project Hosting on Google Code

Older Sep 23, 2009 r1017 (Brian's updates.) committed by brian.t.tanner - Brian's updates. Brian's updates. Sep 23, 2009 r1016 (Added tanner09a.tex, the edited file.) committed by brian.t.tanner - Added tanner09a.tex, the edited file. Added tanner09a.tex, the edited file. Aug 19, 2009 r1015 (Changed editor to Mikio.) committed by brian.t.tanner - Changed editor to Mikio. Changed editor to Mikio. Aug 19, 2009 r1014 (Created the final folder for JMLR.) committed by brian.t.tanner - Created the final folder for JMLR. Created the final folder for JMLR. Aug 19, 2009 r1013 (Last touches before making the final version ) committed by brian.t.tanner - Last touches before making the final version Last touches before making the final version Aug 02, 2009 r1012 (few small changes) committed by whitienator - few small changes few small changes Jul 23, 2009 r1011 (Couple small updates because of review feedback.) committed by brian.t.tanner - Couple small updates because of review feedback. Couple small updates because of review feedback. Jun 29, 2009 r1010 (Small update to the documentation fixing some broken links.) committed by brian.t.tanner - Small update to the documentation fixing some broken links. Small update to the documentation fixing some broken links. Jun 29, 2009 r1009 (Small update to the documentation fixing some broken links.) committed by brian.t.tanner - Small update to the documentation fixing some broken links. Small update to the documentation fixing some broken links. Jun 29, 2009 r1008 (Updated some broken links in the technical manual.) committed by brian.t.tanner - Updated some broken links in the technical manual. Updated some broken links in the technical manual. Jun 06, 2009 issue 128 (Crash with Python Codec if episode terminates on first RL_St...) Status changed by brian.t.tanner - Jason verified that this fixed his problem, so closing. Status: Fixed Jason verified that this fixed his problem, so closing. Status: Fixed Jun 05, 2009 issue 128 (Crash with Python Codec if episode terminates on first RL_St...) commented on by brian.t.tanner - I think this is fixed with r1006. Fixed a bug in the RL_network.c code where it would dereference a null pointer if an empty abstract type was sent into rlCopyADTToBuffer. There was code to handle this, but it was messed up. Pushed out new RL-Glue 3.04. I think this is fixed with r1006. Fixed a bug in the RL_network.c code where it would dereference a null pointer if an empty abstract type was sent into rlCopyADTToBuffer. There was code to handle this, but it was messed up. Pushed out new RL-Glue 3.04. Jun 05, 2009 r1007 (Tried to bootstrap from UBuntu because my new Mac doesn't ha...) committed by brian.t.tanner - Tried to bootstrap from UBuntu because my new Mac doesn't have the right tools. Tried to bootstrap from UBuntu because my new Mac doesn't have the right tools. Jun 05, 2009 r1006 (Updated RL-Glue to 3.04 from 3.03. Fixed a bug in the RL_net...) committed by brian.t.tanner - Updated RL-Glue to 3.04 from 3.03. Fixed a bug in the RL_network.c code where it would dereference a null pointer if an empty abstract type was sent into rlCopyADTToBuffer. There was code to handle this, but it was messed up. Updated RL-Glue to 3.04 from 3.03. Fixed a bug in the RL_network.c code where it would dereference a null pointer if an empty abstract type was sent into rlCopyADTToBuffer. There was code to handle this, but it was messed up. Jun 05, 2009 issue 128 (Crash with Python Codec if episode terminates on first RL_St...) commented on by brian.t.tanner - Tried to replicate the problem in Java. A similar environment written in Java doesn't not crash. Going to peek at the Python codec. Tried to replicate the problem in Java. A similar environment written in Java doesn't not crash. Going to peek at the Python codec. Jun 05, 2009 issue 128 (Crash with Python Codec if episode terminates on first RL_St...) reported by brian.t.tanner - I've attached a simple example from Jason Williams that crashes the latest version of RL-Glue and Python codec. There is one environment that crashes the glue, one doesn't, and the only difference is the crasher ends the episode on the first step. I've attached a simple example from Jason Williams that crashes the latest version of RL-Glue and Python codec. There is one environment that crashes the glue, one doesn't, and the only difference is the crasher ends the episode on the first step. Apr 09, 2009 issue 51 (Porting RL_Glue to true windows) Labels changed by xjamartinh - Edmonton, we had a problem !. At least in my windows installation i cant build the rl_glue.exe it is neccesary to modify manually the file Makefile.in # force to add this libraries -lws2_32 -lkernel32 -luser32 -ladvapi32 -lshell32 LIBS = -lws2_32 -lkernel32 -luser32 -ladvapi32 -lshell32 #LIBS = @LIBS@ without doing this I am unable to make rl-glue Labels: Type-Patch Priority-Critical OpSys-Windows −Type-Enhancement −Priority-High Edmonton, we had a problem !. At least in my windows installation i cant build the rl_glue.exe it is neccesary to modify manually the file Makefile.in # force to add this libraries -lws2_32 -lkernel32 -luser32 -ladvapi32 -lshell32 LIBS = -lws2_32 -lkernel32 -luser32 -ladvapi32 -lshell32 #LIBS = @LIBS@ without doing this I am unable to make rl-glue Labels: Type-Patch Priority-Critical OpSys-Windows −Type-Enhancement −Priority-High Mar 27, 2009 issue 127 (Equality function for actions and observations should do dee...) reported by anna.koop - What steps will reproduce the problem? 1. Make two different action objects, fill them identically 2. a1.equals(a2) returns false What is the expected output? What do you see instead? Expect .equals to look at the content What version of the product are you using? On what operating system? Mac, I'm lazy Please provide any additional information below. What steps will reproduce the problem? 1. Make two different action objects, fill them identically 2. a1.equals(a2) returns false What is the expected output? What do you see instead? Expect .equals to look at the content What version of the product are you using? On what operating system? Mac, I'm lazy Please provide any additional information below. Mar 06, 2009 r1005 (Final version of paper before submission.) committed by br...@tannerpages.com - Final version of paper before submission. Final version of paper before submission. Mar 06, 2009 r1004 (a few tiny typos and extended teh space between references t...) committed by br...@tannerpages.com - a few tiny typos and extended teh space between references to fit exactly 4 pages. a few tiny typos and extended teh space between references to fit exactly 4 pages. Mar 06, 2009 r1003 (Updated size of text in figure, but in older refs for succes...) committed by br...@tannerpages.com - Updated size of text in figure, but in older refs for successes, and cheated in various ways to get to 4 pages. Will fix for final hopefully. Updated size of text in figure, but in older refs for successes, and cheated in various ways to get to 4 pages. Will fix for final hopefully. Mar 05, 2009 r1002 (Moved graffles to figures_raw, added figures directory for p...) committed by br...@tannerpages.com - Moved graffles to figures_raw, added figures directory for pdf figures, made glue.pdf transparent background, implemented a pile of changes that adam and I talked about. Moved graffles to figures_raw, added figures directory for pdf figures, made glue.pdf transparent background, implemented a pile of changes that adam and I talked about. Mar 03, 2009 r1001 (Brian's edits.) committed by br...@tannerpages.com - Brian's edits. Brian's edits. Mar 03, 2009 r1000 (modified rl-glue section) committed by whitienator - modified rl-glue section modified rl-glue section Mar 03, 2009 r999 (modified intro) committed by whitienator - modified intro modified intro Mar 02, 2009 r998 (Some updates.) committed by br...@tannerpages.com - Some updates. Some updates. Feb 27, 2009 r997 (Elliot's concerns have been addressed.) committed by br...@tannerpages.com - Elliot's concerns have been addressed. Elliot's concerns have been addressed. Feb 25, 2009 r996 (Bumped version to 3.03 to test some debian stuff.' ) committed by br...@tannerpages.com - Bumped version to 3.03 to test some debian stuff.' Bumped version to 3.03 to test some debian stuff.' Feb 20, 2009 r995 (Brad and Scott's feedback integrated.) committed by br...@tannerpages.com - Brad and Scott's feedback integrated. Brad and Scott's feedback integrated. Feb 18, 2009 r994 (Small edits from Brian.) committed by br...@tannerpages.com - Small edits from Brian. Small edits from Brian. Feb 18, 2009 r993 (small changes) committed by whitienator - small changes small changes Feb 17, 2009 r992 (All changes from Tuesday meaning, except perhaps thanking fu...) committed by br...@tannerpages.com - All changes from Tuesday meaning, except perhaps thanking funding agencies in acknowledgements. All changes from Tuesday meaning, except perhaps thanking funding agencies in acknowledgements. Feb 15, 2009 r991 (Updated makefile to include the man page.) committed by br...@tannerpages.com - Updated makefile to include the man page. Updated makefile to include the man page. Feb 13, 2009 r990 (Brian made his second pass.) committed by br...@tannerpages.com - Brian made his second pass. Brian made his second pass. Feb 13, 2009 r989 (Brian made his second pass.) committed by br...@tannerpages.com - Brian made his second pass. Brian made his second pass. Feb 13, 2009 r988 (Updated RL_network.c to use local sockets.) committed by br...@tannerpages.com - Updated RL_network.c to use local sockets. Updated RL_network.c to use local sockets. Feb 13, 2009 r987 (Branching to see how much faster local sockets are than netw...) committed by br...@tannerpages.com - Branching to see how much faster local sockets are than network sockets. Branching to see how much faster local sockets are than network sockets. Feb 13, 2009 r986 (A couple of typos in the docs, and bumped the version to 3.0...) committed by br...@tannerpages.com - A couple of typos in the docs, and bumped the version to 3.01 (because of the man page). A couple of typos in the docs, and bumped the version to 3.01 (because of the man page). Feb 12, 2009 r985 (Brian has made 1 pass through the document, some things for ...) committed by br...@tannerpages.com - Brian has made 1 pass through the document, some things for the better, some for the worse. Brian has made 1 pass through the document, some things for the better, some for the worse. Feb 12, 2009 issue 104 (Get specific people to try out each codec) Status changed by br...@tannerpages.com - This went great! Status: Fixed This went great! Status: Fixed Feb 12, 2009 issue 65 (Unix Testing) changed by br...@tannerpages.com - It's well tested. Status: Fixed Owner: br...@tannerpages.com It's well tested. Status: Fixed Owner: br...@tannerpages.com Feb 12, 2009 issue 124 (RL-Glue 3.0 Final Release Candidate Code Review) Status changed by br...@tannerpages.com - Thanks for the feedback, I think we addressed almost all of it. Status: Fixed Thanks for the feedback, I think we addressed almost all of it. Status: Fixed Feb 12, 2009 r984 (RL-Glue, now with man page\!) committed by br...@tannerpages.com - RL-Glue, now with man page\! RL-Glue, now with man page\! Feb 12, 2009 r983 (Added man page for RL-Glue.) committed by br...@tannerpages.com - Added man page for RL-Glue. Added man page for RL-Glue. Feb 10, 2009 r982 (a few small changes) committed by whitienator - a few small changes a few small changes Feb 10, 2009 issue 51 (Porting RL_Glue to true windows) commented on by xjamartinh - As a Windows user I think there are at least 3 ways in which standalone RLGlue (i.e wihtout cygwin) for windows should be distributed. 1.- precompiled binary package with an automatic installer. 2.- source code and a make file in the MinGW format. 3.- source code and a visual studio project. It is possible however to build RL-Glue.exe without cygwin using MSYS but altoutgh this produces a standalone rlglue.exe (no cygwin dll dependency) it is based on a linux terminal emulation with is not the expected and indeed is a mostly unkown way of building software on windows. Windows users are in general dummies, they (or we) expect things to be very just as simple as getting an installer, executing it and then find under the menu the program we have installed, just as matlab or another development tool. I know that maybe producing such a distribution for an open source development tool such as RLGlue which is for research is difficult and may seems to be unecessary but it is not. So, I think that with the three mentioned possible distributions for windows it is ok. Indeed the precompiled binaries could be left for another person wich in general distribute pre-built binaries as i usuall on the windows community. So maybe the most easy is to produce a basic MinGW bulding system which will only requeire the download and installation of MinGW. hope this contribute to give some lights about what windows users expect from opensoure software. As a Windows user I think there are at least 3 ways in which standalone RLGlue (i.e wihtout cygwin) for windows should be distributed. 1.- precompiled binary package with an automatic installer. 2.- source code and a make file in the MinGW format. 3.- source code and a visual studio project. It is possible however to build RL-Glue.exe without cygwin using MSYS but altoutgh this produces a standalone rlglue.exe (no cygwin dll dependency) it is based on a linux terminal emulation with is not the expected and indeed is a mostly unkown way of building software on windows. Windows users are in general dummies, they (or we) expect things to be very just as simple as getting an installer, executing it and then find under the menu the program we have installed, just as matlab or another development tool. I know that maybe producing such a distribution for an open source development tool such as RLGlue which is for research is difficult and may seems to be unecessary but it is not. So, I think that with the three mentioned possible distributions for windows it is ok. Indeed the precompiled binaries could be left for another person wich in general distribute pre-built binaries as i usuall on the windows community. So maybe the most easy is to produce a basic MinGW bulding system which will only requeire the download and installation of MinGW. hope this contribute to give some lights about what windows users expect from opensoure software. Feb 09, 2009 r981 (Fixed up makefile to bring readme.txt with distribution.) committed by br...@tannerpages.com - Fixed up makefile to bring readme.txt with distribution. Fixed up makefile to bring readme.txt with distribution. Feb 09, 2009 issue 126 (Build warnings on Linux.) changed by br...@tannerpages.com - I added #include<strings.h> in the task spec parser. this got rid of the warning on linux, nothing changed on Mac. Windows now has a warning (maybe always did). On windows, the mines example that does task spec decoding seems to run, so I'm assuming life is good. Status: Fixed Cc: ScottCLivingston I added #include<strings.h> in the task spec parser. this got rid of the warning on linux, nothing changed on Mac. Windows now has a warning (maybe always did). On windows, the mines example that does task spec decoding seems to run, so I'm assuming life is good. Status: Fixed Cc: ScottCLivingston Feb 09, 2009 r980 (Committing RL-Glue-3.0.) committed by br...@tannerpages.com - Committing RL-Glue-3.0. Committing RL-Glue-3.0. Feb 09, 2009 issue 125 (rl_glue.exe hard to find on Windows) commented on by br...@tannerpages.com - Oops. I totally edited the wrong issue. Sorry. Disregard last comment. Oops. I totally edited the wrong issue. Sorry. Disregard last comment.

issue 51 (Porting RL_Glue to true windows) commented on by xjamartinh - As a Windows user I think there are at least 3 ways in which standalone RLGlue (i.e wihtout cygwin) for windows should be distributed. 1.- precompiled binary package with an automatic installer. 2.- source code and a make file in the MinGW format. 3.- source code and a visual studio project. It is possible however to build RL-Glue.exe without cygwin using MSYS but altoutgh this produces a standalone rlglue.exe (no cygwin dll dependency) it is based on a linux terminal emulation with is not the expected and indeed is a mostly unkown way of building software on windows. Windows users are in general dummies, they (or we) expect things to be very just as simple as getting an installer, executing it and then find under the menu the program we have installed, just as matlab or another development tool. I know that maybe producing such a distribution for an open source development tool such as RLGlue which is for research is difficult and may seems to be unecessary but it is not. So, I think that with the three mentioned possible distributions for windows it is ok. Indeed the precompiled binaries could be left for another person wich in general distribute pre-built binaries as i usuall on the windows community. So maybe the most easy is to produce a basic MinGW bulding system which will only requeire the download and installation of MinGW. hope this contribute to give some lights about what windows users expect from opensoure software.

As a Windows user I think there are at least 3 ways in which standalone RLGlue (i.e wihtout cygwin) for windows should be distributed. 1.- precompiled binary package with an automatic installer. 2.- source code and a make file in the MinGW format. 3.- source code and a visual studio project. It is possible however to build RL-Glue.exe without cygwin using MSYS but altoutgh this produces a standalone rlglue.exe (no cygwin dll dependency) it is based on a linux terminal emulation with is not the expected and indeed is a mostly unkown way of building software on windows. Windows users are in general dummies, they (or we) expect things to be very just as simple as getting an installer, executing it and then find under the menu the program we have installed, just as matlab or another development tool. I know that maybe producing such a distribution for an open source development tool such as RLGlue which is for research is difficult and may seems to be unecessary but it is not. So, I think that with the three mentioned possible distributions for windows it is ok. Indeed the precompiled binaries could be left for another person wich in general distribute pre-built binaries as i usuall on the windows community. So maybe the most easy is to produce a basic MinGW bulding system which will only requeire the download and installation of MinGW. hope this contribute to give some lights about what windows users expect from opensoure software.

Older