freenode/#sbcl - IRC Chatlog

23:40:24 stassats unlike C, you don't have to recompile the whole thing and restart to insert a print or a break

23:41:08 stassats oh, and TRACE

23:44:42 lottaquestions_ Yeah, I've heard of trace. I am still in the discovery phase of discovering the limits of sbcl and slime

23:45:20 lottaquestions_ there have been cases where I recompiled and loaded, and sbcl ignored newly added break forms

23:45:54 stassats it didn't ignore, you just didn't recompile properly

23:46:13 stassats or comprehensively

23:46:35 lottaquestions_ ok....that I would like to know how to do :-)

23:46:45 lottaquestions_ it would save me a lot of headache

23:47:02 Bike_ ** NICK Bike

23:47:49 lottaquestions_ what I have been doing has been C-c C-k for compile and load the current buffer's file

23:48:11 lottaquestions_ what have I been missing to make it a comprehensive recompile?

23:48:38 lottaquestions_ this is using SBCL via slime

23:49:05 stassats not calling the new code

23:51:11 lottaquestions_ To call the new code, I was restarting the frame above the frame that contains the new code

23:51:50 lottaquestions_ from my understanding this would call the new code with my changes.

23:52:04 stassats depends on what the code does

23:52:09 lottaquestions_ unless of course there is something different I should be doing to run the new code

23:53:20 stassats and restarting frames doesn't always get the arguments right, so be mindful of that

23:53:36 lottaquestions_ yes, I already faced that....

23:54:00 lottaquestions_ Even had a case where the global vars disappeared

23:54:09 stassats compiling new code doesn't change the actual code, just the name association to that code

23:54:22 lottaquestions_ after restarting a frame that I thought was sufficiently deep in the stack

23:55:51 lottaquestions_ basically all I have had as pointers on how to do this is a video by a one Marco Baringer, a blog by someone with the handle malisper and a video of some other guy debugging a web server (hunchentoot maybe? not sure)

23:56:50 lottaquestions_ I wouldn't mind pointers on the subtleties of SBCL+slime to allow me to use it effectively

23:58:07 lottaquestions_ oh...<compiling new code doesn't change the actual code, just the name association to that code>

23:58:50 lottaquestions_ how do I change the actual code ? I do not mind going and reading up if there is documentation somewhere that actually explains this

23:59:20 stassats well, you can't

23:59:24 stassats and you shouldn't want to

0:00:09 lottaquestions_ ok,...how do I correctly do a hot compile and reload

0:00:18 stassats you have to stop using any old code and start calling the new code

0:00:36 stassats and how to do that... depends on the code

1:04:59 |3b| lottaquestions_: if you didn't try it already, higher DEBUG (with local declarations, restrict-compiler-policy, or C-u C-c C-k, C-u C-c C-c etc in slime) might give you better debugging

2:02:15 lottaquestions_ @|3b| I have this set at the beginning of all my files: (declaim (optimize (debug 3)))

2:02:30 lottaquestions_ I guess that is what you mean by higher debug

2:10:50 |3b| yeah, that probably works

2:11:44 |3b| (though arguably bad style for any code you share with others, since it affects everything loaded after it too)

2:13:51 lottaquestions_ That's right. One thing that really puzzled me about what stassats said was that one has to stop using the old code and start calling new code.

2:14:35 |3b| right, recompiling just generates new code, and (probably) binds it to some name

2:15:55 |3b| it has no effect on existing code, so a running function won't be affected by recompiling that function. only calls made after the binding is changed will see the new code

2:15:57 lottaquestions_ so say I have a function called foo, that I recompile with C-u C-c C-c, how do I get to use the new version of foo instead of the old one? stassats seemed to be saying that there is a way of doing it

2:16:18 |3b| just call it

2:16:47 lottaquestions_ Exactly what I thought

2:18:25 |3b| just not always obvious what counts as "call it" in that case, for example inlined code doesn't get automatically recompiled, and function objects (for example returned from #' ) only see the specific code from when it was evaluated, not any code associated with that name later

2:20:17 |3b| and code interrupted by debugger/stepper/etc continues executing the old code, though sounds like you got that part. restarting something higher on the stack usually works for that (when it is calling through the binding)

2:23:01 lottaquestions_ I already experienced the function objects issue :-), and had interesting results with restarting frames ie disappearing global variables

2:23:57 |3b| yeah, debugger isn't perfect, so sometimes you have to just exit and start from top call

2:24:53 lottaquestions_ thanks for the pointers

2:25:17 lottaquestions_ at least now I know I am going in the right direction :-)

2:25:21 |3b| restarting frames doesn't help much when there is external state (files, network, destructive modification of data structures, etc), since nothing restores that state

2:26:12 |3b| (unless you write a bunch of magic unwind-protect code to simulate transactions or something, but even that probably wouldn't help with network code)