1. memory leaks and colors
- Posted by useless_ Feb 01, 2013
- 1840 views
so which version of Eu has 1) the match/replace memory leak fixed, 2) the debugger colors fixed? 3) the dll memory leak fixed?
useless
2. Re: memory leaks and colors
- Posted by useless_ Feb 08, 2013
- 1809 views
I'd still like an answer to my question. In addition, i offer this screen capture:
http://designerthinking.com/1297.GIF
Note line 1297 in the debug window, which is highlited in the text editor.
I have had to break several programs into pieces and call them with cmdline args repeatedly to avoid one long run crashing due to memory leaks.
useless
3. Re: memory leaks and colors
- Posted by useless_ Feb 09, 2013
- 1692 views
Still awaiting a heads up for the version of Eu that has the leaks patched and the debugger colors fixed.
useless
4. Re: memory leaks and colors
- Posted by mattlewis (admin) Feb 09, 2013
- 1662 views
Still awaiting a heads up for the version of Eu that has the leaks patched and the debugger colors fixed.
I'm not aware of any leaks in the current tips of either the 4.0 or 4.1 code (likewise, I've been told, though haven't personally verified that the debugger colors have been fixed). You've mentioned that you're still seeing some, but I have no idea where they're coming from. Do you have anything fairly minimal that reproduces it? If not, even the full thing would be helpful.
Matt
5. Re: memory leaks and colors
- Posted by useless_ Feb 09, 2013
- 1655 views
Still awaiting a heads up for the version of Eu that has the leaks patched and the debugger colors fixed.
I'm not aware of any leaks in the current tips of either the 4.0 or 4.1 code (likewise, I've been told, though haven't personally verified that the debugger colors have been fixed). You've mentioned that you're still seeing some, but I have no idea where they're coming from. Do you have anything fairly minimal that reproduces it? If not, even the full thing would be helpful.
Matt
.. 3 am Friday 2/1/2013 .. [03:02] <useless2> other run-away memory leak happening [03:04] <useless2> i am heavily using match(), match_from(), match_all() [03:04] <useless2> memory use is bouncing from 80kb to 1.5Gb [03:05] <useless2> it's pushed pf use up to 4.5Gbytes [03:06] <useless2> just crashed : couldn't alloc 482156 bytes<snip>
.. 1 pm Friday 2/1/2013 .. [13:36] <useless2> using Euphoria-4.0.5-Jan30-2013 [13:36] <useless2> couldn't alloc 55200 bytes [13:36] <useless2> Your program has run out of memory. [13:36] <useless2> One moment please... .. 2 pm Friday 2/1/2013 .. [14:40] <useless2> ...We discussed it on irc 8 days ago:
.. 3 pm Friday 2/1/2013 .. [15:04] <mattlewis> hmm...that one should have the replace() memory leak fixed [15:05] <useless2> i agree [15:06] <mattlewis> I presume that your program releases the data? ...and this looks like another leak [15:06] <useless2> yeas, here's very few vars in use, each iteration assigns a new value (rather than appending)
Since it has been a week since i first asked about it, and you saying it should have been fixed, and now saying it is fixed, i didn't skinny the code down because it's plain you aren't interested in working with me on the problem.
If all leaks ARE fixed in the current release, what windows release is that? And is the color-fixed debug-trace included?
useless
6. Re: memory leaks and colors
- Posted by useless_ Feb 10, 2013
- 1659 views
[00:50] <useless2> i pasted a test file for the memory leak, and the Eu website stripped it bare so it is useless
[00:51] <useless2> but it's run on a eui version that's KNOWN to have a memory leak, because in the WEEK since i have mentioned a problem, NO ONE HAS TOLD ME WHICH EUI VERSION HAS BEEN FIXED AND ACTUALLY RUNS .. 1 am Sunday 2/10/2013 ..
useless
7. Re: memory leaks and colors
- Posted by mattlewis (admin) Feb 10, 2013
- 1626 views
Since it has been a week since i first asked about it, and you saying it should have been fixed, and now saying it is fixed, i didn't skinny the code down because it's plain you aren't interested in working with me on the problem.
If all leaks ARE fixed in the current release, what windows release is that? And is the color-fixed debug-trace included?
As far as I know, the leaks have been fixed. I believe that the problem with the nightly builds has been fixed, so if you grab the most recent version from there (of whichever branch you want to use) you should be leak free. If not, there there is an additional leak that I'm not aware of.
Matt
8. Re: memory leaks and colors
- Posted by useless_ Feb 10, 2013
- 1603 views
Since it has been a week since i first asked about it, and you saying it should have been fixed, and now saying it is fixed, i didn't skinny the code down because it's plain you aren't interested in working with me on the problem.
If all leaks ARE fixed in the current release, what windows release is that? And is the color-fixed debug-trace included?
As far as I know, the leaks have been fixed. I believe that the problem with the nightly builds has been fixed, so if you grab the most recent version from there (of whichever branch you want to use) you should be leak free. If not, there there is an additional leak that I'm not aware of.
Matt
I have repeatedly tried to make you aware of it.
See: http://openeuphoria.org/pastey/195.wc
useless
9. Re: memory leaks and colors
- Posted by mattlewis (admin) Feb 10, 2013
- 1621 views
Since it has been a week since i first asked about it, and you saying it should have been fixed, and now saying it is fixed, i didn't skinny the code down because it's plain you aren't interested in working with me on the problem.
If all leaks ARE fixed in the current release, what windows release is that? And is the color-fixed debug-trace included?
As far as I know, the leaks have been fixed. I believe that the problem with the nightly builds has been fixed, so if you grab the most recent version from there (of whichever branch you want to use) you should be leak free. If not, there there is an additional leak that I'm not aware of.
Matt
I have repeatedly tried to make you aware of it.
See: http://openeuphoria.org/pastey/195.wc
useless
Yup, that's a new leak. 
Matt
10. Re: memory leaks and colors
- Posted by useless_ Feb 10, 2013
- 1664 views
Since it has been a week since i first asked about it, and you saying it should have been fixed, and now saying it is fixed, i didn't skinny the code down because it's plain you aren't interested in working with me on the problem.
If all leaks ARE fixed in the current release, what windows release is that? And is the color-fixed debug-trace included?
As far as I know, the leaks have been fixed. I believe that the problem with the nightly builds has been fixed, so if you grab the most recent version from there (of whichever branch you want to use) you should be leak free. If not, there there is an additional leak that I'm not aware of.
Matt
I have repeatedly tried to make you aware of it.
See: http://openeuphoria.org/pastey/195.wc
useless
Yup, that's a new leak.
Matt
Please announce when this bug is fixed.
useless
11. Re: memory leaks and colors
- Posted by SDPringle Feb 10, 2013
- 1623 views
These kinds of things can be avoided largely by writing the majority of code in EUPHORIA, rather than trying to make every thing that can be done in EUPHORIA as a builtin or via C-call/interface. Although the lack of former structures is a big problem for using accidentally the wrong member value offset in the wrong structure.
Shawn Pringle
12. Re: memory leaks and colors
- Posted by mattlewis (admin) Feb 10, 2013
- 1625 views
Please announce when this bug is fixed.
ticket:847 has been fixed. The next nightly builds should pick up the fix.
Matt
13. Re: memory leaks and colors
- Posted by mattlewis (admin) Feb 10, 2013
- 1600 views
These kinds of things can be avoided largely by writing the majority of code in EUPHORIA, rather than trying to make every thing that can be done in EUPHORIA as a builtin or via C-call/interface. Although the lack of former structures is a big problem for using accidentally the wrong member value offset in the wrong structure.
That's true, but the new build-in sequence functions can result in really big performance gains vs euphoria-only solutions.
Matt
14. Re: memory leaks and colors
- Posted by SDPringle Feb 10, 2013
- 1602 views
Yes certainly a routine written in C can be expected to be faster. However, people use a language devoid of pointers to avoid debugging dangling pointers and memory leaks. I sure hope there are none of the former. I would rather have memory leaks 99.9% of the time than dangling pointers 0.1% of the time!
We should keep the 4.0 branch (bug fix only) going long after 4.1 is released in light of all of these bugs.
Shawn Pringle
15. Re: memory leaks and colors
- Posted by EUWX Feb 10, 2013
- 1537 views
We should keep the 4.0 branch (bug fix only) going long after 4.1 is released in light of all of these bugs.
Shawn Pringle
I second the motion!
16. Re: memory leaks and colors
- Posted by useless_ Feb 11, 2013
- 1571 views
Please announce when this bug is fixed.
ticket:847 has been fixed. The next nightly builds should pick up the fix.
Matt
The bug is indeed squashed in eubins-watcom-20130210-42b209901e89.zip. And the debugger has the correct colors again too. Thanks.
useless
17. Re: memory leaks and colors
- Posted by mattlewis (admin) Feb 11, 2013
- 1563 views
Yes certainly a routine written in C can be expected to be faster. However, people use a language devoid of pointers to avoid debugging dangling pointers and memory leaks. I sure hope there are none of the former. I would rather have memory leaks 99.9% of the time than dangling pointers 0.1% of the time!
OK. I'm not sure what this all has to do with the topic at hand. I doubt you'd really want to have so many memory leaks.
We should keep the 4.0 branch (bug fix only) going long after 4.1 is released in light of all of these bugs.
Of course, anyone can work on that if they want (hey, people are still hacking on the 3.1 stuff!). Personally, I'm not planning on doing anything in 4.0 once 4.1 is out.
Matt
18. Re: memory leaks and colors
- Posted by jimcbrown (admin) Feb 11, 2013
- 1591 views
We should keep the 4.0 branch (bug fix only) going long after 4.1 is released in light of all of these bugs.
Of course, anyone can work on that if they want (hey, people are still hacking on the 3.1 stuff!). Personally, I'm not planning on doing anything in 4.0 once 4.1 is out.
Matt
I feel the dev team should drop 4.0 support after 4.0.6 - we're stretched pretty thin as it is. This also means more eyes on the 4.1 branch, and thus a better chance of caching more bugs.
Also, I was thinking a tiered release of 4.1 like we did for 4.0.0 - an alpha or two, a couple of beta release, and then some release canidates.
19. Re: memory leaks and colors
- Posted by mattlewis (admin) Feb 11, 2013
- 1641 views
Please announce when this bug is fixed.
ticket:847 has been fixed. The next nightly builds should pick up the fix.
Matt
The bug is indeed squashed in eubins-watcom-20130210-42b209901e89.zip. And the debugger has the correct colors again too. Thanks.
I did some digging. That bug has been in there since its initial implementation in October 2010. Thanks for uncovering it.
Matt
20. Re: memory leaks and colors
- Posted by useless_ Feb 12, 2013
- 1750 views
Please announce when this bug is fixed.
ticket:847 has been fixed. The next nightly builds should pick up the fix.
Matt
The bug is indeed squashed in eubins-watcom-20130210-42b209901e89.zip. And the debugger has the correct colors again too. Thanks.
I did some digging. That bug has been in there since its initial implementation in October 2010. Thanks for uncovering it.
Matt
This is wonderful, the app has processed 6 gigabytes of data so far, and still using only ~5 megabytes of memory. This is how it always should have been! (I mean other than the 6300 bytes of source code using 4 megabytes of memory.)
Thanks, Matt.
useless
