Search
Monday, 30th of March 2020, 0:44:08 UTC
3:02:02
beach
Good morning everyone!
12:19:58
kpoeck
drmeister I am about the integrate the regression-tests in waf as `./waf regression_tests`
12:20:25
kpoeck
Can I rely on "build/clasp being present?
12:21:28
drmeister
For now I think you would have to. Unless you can figure out how to require waf to build it if it isn't there.
12:21:37
kpoeck
i think it is no longer created for `USE_COMPILE_FILE_PARALLEL=False`I 'd fix that too
12:21:48
Bike
hm... i'd say use iclasp-boehm, but then you're assuming boehm build
12:21:57
drmeister
./waf analyze_clasp depends on build/boehm/iclasp-boehm and build/mpsprep
12:22:13
Bike
can we like, always make it? like even if there's whatever obscure linking thing, make a shell script instead of an executable there
12:22:18
drmeister
Rather it relies on them being present.
12:22:41
drmeister
Could you guys figure out how to do that? Check the waf book?
12:22:48
drmeister
I'd like to know myself.
12:22:53
kpoeck
After every build i do "ln -s -f ~/lisp/compiler/clasp-karsten/build/boehm/cclasp-boehm ~/lisp/compiler/clasp-karsten/build/clasp"
12:23:14
kpoeck
I think I can repair wscript so that it is done automatically
12:23:32
Bike
well we don't want a link if there's an actual executable, right?
12:28:35
kpoeck
I also looked at signal-handling and sigfpe, please have a look on https://github.com/clasp-developers/clasp/pull/955
12:30:20
Bike
something with no obvious documentation. super.
12:31:28
Bike
the out of extent probably has nothing to do with your change
12:32:21
Bike
honestly, returning out of a signal handler is just kinda dicey
12:32:30
Bike
should be possible, but dicey
12:36:04
kpoeck
#if !defined(_POSIX_C_SOURCE) || defined(_DARWIN_C_SOURCE)#define FPE_NOOP 0 /* if only I knew... */#endif
12:36:14
Bike
is that the actual comment
12:37:01
Bike
oh, why did you turn off init_float_traps()? when i did t hat before i hit problems when i did more than one FPU error, i think
12:37:12
Bike
of course, if you can't recover from an FPU error to begin with, i guess it doesn't matter
12:37:20
Bike
(but not being able to do so sucks)
12:40:05
kpoeck
I believe init_float_traps i a no opt since CLASP_FPT_SUPPORT is false
12:40:52
kpoeck
But why would I call this from handle_fpe?
12:41:14
Bike
mm, then it would be a nop, yeah
12:41:23
Bike
i had it calling that because entering the signal handler seemed to reset them
12:41:45
kpoeck
Well thinking harder I believe this is just the case on darwin, might work on linux
12:42:43
kpoeck
So perhaps I better add it again, what is your opinion?
12:43:40
Bike
mm, well if it's a no-op anyway having it there is fairly harmless, and it might eventually not be a no-op. but we can probably just leave it out until we get more of a handle on things, by which i mean fixing the out of extent return
Monday, 30th of March 2020, 12:44:08 UTC