Re: [sage-devel] [Debian] Strange crash in ECL
On 26/02/2015, at 20:57, Julien Puydt julien.pu...@laposte.net wrote: Le 26/02/2015 08:20, Francois Bissey a écrit : On 26/02/2015, at 20:12, Julien Puydt julien.pu...@laposte.net wrote: Le 25/02/2015 22:43, François Bissey a écrit : That's fun, on Gentoo my ecl from the system just chew it without crashing. What options are you using to build ecl I currently have --enable-unicode -with-x and the rest turned off. Would it be thread related by any chance? Sage is using mpir and doesn't have the problem. Debian is using gmp and does have the problem. What are you using? Oh, gmp just as you do. Using mpir is hard on sage-on-distro as you would know. Snif. The gmp vs mpir difference was my favored explanation after I ruled out : - disable-threads for ecl - enable-large-config for Boehm's gc Debian and upstream-on-debian have the crash. I don't know what the upstream developer was using when he reproduced it. Gentoo doesn't have the crash, and I had a report ubuntu didn't either. Ok what version of gmp? 5.1.3 here. François -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
On 26/02/2015, at 21:47, Julien Puydt julien.pu...@laposte.net wrote: Le 26/02/2015 09:05, Francois Bissey a écrit : Ok what version of gmp? 5.1.3 here. I have 6.0.0. OK upgraded to 6.0.0a (the one I have available in the tree): (setq b (expt a 600)) Condition of type: SEGMENTATION-VIOLATION Detected access to an invalid or protected memory address. Available restarts: 1. (CONTINUE) Ignore signal 2. (RESTART-TOPLEVEL) Go back to Top-Level REPL. Broken at SI:BYTECODES. [Evaluation of: (SETQ B …)] So I think we can say there is something going with gmp 6.0.0. I think that should narrow things down considerably. François -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
Le 26/02/2015 10:08, Francois Bissey a écrit : So I think we can say there is something going with gmp 6.0.0. I think that should narrow things down considerably. I'll report our findings to upstream -- that should definitely interest them, and help pinpoint the issue. Thanks! Snark on #sagemath -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
Le 26/02/2015 10:15, Julien Puydt a écrit : Le 26/02/2015 10:08, Francois Bissey a écrit : So I think we can say there is something going with gmp 6.0.0. I think that should narrow things down considerably. I'll report our findings to upstream -- that should definitely interest them, and help pinpoint the issue. I set ECL_OPT_SET_GMP_MEMORY_FUNCTIONS to 0 in src/c/main.d, and that made the crash disappear! Snark on #sagemath -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
Le 26/02/2015 09:05, Francois Bissey a écrit : Ok what version of gmp? 5.1.3 here. I have 6.0.0. Snark -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
Hi, Le 25/02/2015 10:56, Julien Puydt a écrit : it is a strange problem I have : using either debian's ecl or upstream's the following code : (ext:set-limit 'ext:heap-size 0) (setq a (expt 10 (expt 10 5))) (setq b (expt a 600)) gives a crash (SIGABRT) Duplicate large block deallocation. Where things get interesting, is that I get this code from sage, where it actually works! What is annoying is that the last line comes from this test (src/sage/interfaces/maxima.py, around line 570) : Test that we can use more than 256MB RAM (see trac :trac:`6722`):: sage: a = maxima(10)^(10^5) sage: b = a^600 # long time -- about 10-15 seconds but trac's ticket #6722 is about spellchecking some modules... so I guess there used to be a problem like the one I have, but I don't know what was done to fix it. Does someone remember ? Upstream acknowledged it is a bug, still present in the devel version. I'll try to help them pinpoint and fix the problem -- if someone here has a clue why it works in sage, I'm sure they would appreciate the information. Snark on #sagemath -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
On 02/26/15 10:30, Julien Puydt wrote: Hi, Le 25/02/2015 10:56, Julien Puydt a écrit : it is a strange problem I have : using either debian's ecl or upstream's the following code : (ext:set-limit 'ext:heap-size 0) (setq a (expt 10 (expt 10 5))) (setq b (expt a 600)) gives a crash (SIGABRT) Duplicate large block deallocation. Where things get interesting, is that I get this code from sage, where it actually works! What is annoying is that the last line comes from this test (src/sage/interfaces/maxima.py, around line 570) : Test that we can use more than 256MB RAM (see trac :trac:`6722`):: sage: a = maxima(10)^(10^5) sage: b = a^600 # long time -- about 10-15 seconds but trac's ticket #6722 is about spellchecking some modules... so I guess there used to be a problem like the one I have, but I don't know what was done to fix it. Does someone remember ? Upstream acknowledged it is a bug, still present in the devel version. I'll try to help them pinpoint and fix the problem -- if someone here has a clue why it works in sage, I'm sure they would appreciate the information. That's fun, on Gentoo my ecl from the system just chew it without crashing. What options are you using to build ecl I currently have --enable-unicode -with-x and the rest turned off. Would it be thread related by any chance? Francois -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
Not really related but it seems ECL saw an update! On Wednesday, February 25, 2015 at 10:43:28 PM UTC+1, François wrote: On 02/26/15 10:30, Julien Puydt wrote: Hi, Le 25/02/2015 10:56, Julien Puydt a écrit : it is a strange problem I have : using either debian's ecl or upstream's the following code : (ext:set-limit 'ext:heap-size 0) (setq a (expt 10 (expt 10 5))) (setq b (expt a 600)) gives a crash (SIGABRT) Duplicate large block deallocation. Where things get interesting, is that I get this code from sage, where it actually works! What is annoying is that the last line comes from this test (src/sage/interfaces/maxima.py, around line 570) : Test that we can use more than 256MB RAM (see trac :trac:`6722`):: sage: a = maxima(10)^(10^5) sage: b = a^600 # long time -- about 10-15 seconds but trac's ticket #6722 is about spellchecking some modules... so I guess there used to be a problem like the one I have, but I don't know what was done to fix it. Does someone remember ? Upstream acknowledged it is a bug, still present in the devel version. I'll try to help them pinpoint and fix the problem -- if someone here has a clue why it works in sage, I'm sure they would appreciate the information. That's fun, on Gentoo my ecl from the system just chew it without crashing. What options are you using to build ecl I currently have --enable-unicode -with-x and the rest turned off. Would it be thread related by any chance? Francois -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
Le 25/02/2015 22:43, François Bissey a écrit : That's fun, on Gentoo my ecl from the system just chew it without crashing. What options are you using to build ecl I currently have --enable-unicode -with-x and the rest turned off. Would it be thread related by any chance? Sage is using mpir and doesn't have the problem. Debian is using gmp and does have the problem. What are you using? Snark on #sagemath -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
On 26/02/2015, at 20:12, Julien Puydt julien.pu...@laposte.net wrote: Le 25/02/2015 22:43, François Bissey a écrit : That's fun, on Gentoo my ecl from the system just chew it without crashing. What options are you using to build ecl I currently have --enable-unicode -with-x and the rest turned off. Would it be thread related by any chance? Sage is using mpir and doesn't have the problem. Debian is using gmp and does have the problem. What are you using? Oh, gmp just as you do. Using mpir is hard on sage-on-distro as you would know. François -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
On 26/02/2015, at 20:14, Julien Puydt julien.pu...@laposte.net wrote: Salut, Le 25/02/2015 23:00, Jean-Pierre Flori a écrit : Not really related but it seems ECL saw an update! Their homepage says 15.2.21, but the dowload page is always 13.5.1. As I mentioned, upstream confirmed the bug, including for their latest and greatest. Someone already put it in the Gentoo tree but I hadn’t noticed before this thread. I haven’t given it a shot yet. François -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
Salut, Le 25/02/2015 23:00, Jean-Pierre Flori a écrit : Not really related but it seems ECL saw an update! Their homepage says 15.2.21, but the dowload page is always 13.5.1. As I mentioned, upstream confirmed the bug, including for their latest and greatest. Snark on #sagemath -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] [Debian] Strange crash in ECL
Le 26/02/2015 08:20, Francois Bissey a écrit : On 26/02/2015, at 20:12, Julien Puydt julien.pu...@laposte.net wrote: Le 25/02/2015 22:43, François Bissey a écrit : That's fun, on Gentoo my ecl from the system just chew it without crashing. What options are you using to build ecl I currently have --enable-unicode -with-x and the rest turned off. Would it be thread related by any chance? Sage is using mpir and doesn't have the problem. Debian is using gmp and does have the problem. What are you using? Oh, gmp just as you do. Using mpir is hard on sage-on-distro as you would know. Snif. The gmp vs mpir difference was my favored explanation after I ruled out : - disable-threads for ecl - enable-large-config for Boehm's gc Debian and upstream-on-debian have the crash. I don't know what the upstream developer was using when he reproduced it. Gentoo doesn't have the crash, and I had a report ubuntu didn't either. Snark on #sagemath -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
[sage-devel] [Debian] Strange crash in ECL
Hi, it is a strange problem I have : using either debian's ecl or upstream's the following code : (ext:set-limit 'ext:heap-size 0) (setq a (expt 10 (expt 10 5))) (setq b (expt a 600)) gives a crash (SIGABRT) Duplicate large block deallocation. Where things get interesting, is that I get this code from sage, where it actually works! What is annoying is that the last line comes from this test (src/sage/interfaces/maxima.py, around line 570) : Test that we can use more than 256MB RAM (see trac :trac:`6722`):: sage: a = maxima(10)^(10^5) sage: b = a^600 # long time -- about 10-15 seconds but trac's ticket #6722 is about spellchecking some modules... so I guess there used to be a problem like the one I have, but I don't know what was done to fix it. Does someone remember ? Snark on #sagemath -- You received this message because you are subscribed to the Google Groups sage-devel group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
