freenode/#clasp - IRC Chatlog
Search
23:48:52
karlosz
drmeister: do you have timings for clasp self compiling cleavir? i wanted to compare my self host time with clasps but it up all my ram again
0:24:25
balrog
drmeister: looks like Ubuntu doesn't package the Clang libs in a way that clasp's wscript expects
0:26:54
drmeister
Because on supercomputers I need static linking - so I started from there and never changed it.
0:29:29
drmeister
We can switch to dynamic linking - but waf seems to want one or the other. Take a look at the wscript where it searches for the boehm library. It tries first to find a static version and then it drops down to the dynamic version.
0:30:33
drmeister
balrog: I don't want to touch another build system until we have loads of developers and there is a really, really, really good reason.
0:31:48
balrog
this would not be the only project I've had to deal with that uses a fairly uncommon build system. :|
0:33:07
drmeister
It's python - I understand python. I don't understand cmake or autotools and I despise boost::build and perforce jam.
0:34:29
balrog
the other build system I'm citing above is a fork of Premake and interestingly enough, uses Lua
0:37:21
drmeister
It's Python and you have function calls to build a dependency graph. We don't use waf the way it was intended - or we barely do.
0:37:59
drmeister
I know a guy who left his job partly because he hated the build system that they decided to switch to.
0:38:35
balrog
I know multiple people who wrote their own build systems because either they hated everything that existed or everything that existed was unsuitable for the application.
0:39:02
drmeister
Two? Three? years ago I switched to waf with no idea how it would go because I was so frustrated with boost::build.
0:39:38
balrog
I've used scons before (also python) and my opinion on it is that it's clean and workable bu tit doesn't scale.
0:40:02
drmeister
All I ever hear is programmers vowing that they are going to develop their own build system.
0:40:38
drmeister
Ha! Where they should really be focusing their efforts is to develop their own compiler.
0:42:16
balrog
haha funny you should say that. one of them is developing FPGA tools, another is developing cross platform compatibility environments to run legacy Apple/Mac software on modern platforms, as well as Unix-like environment on legacy platforms
0:45:54
balrog
just like the effort to improve workflows for computational chemistry led to building a Common Lisp compiler
0:48:58
balrog
my experience with scons? it couldn't handle macOS "fat binaries", it couldn't handle cross compiling
5:42:02
drmeister
::notify Bike In the new dev - can you give me a switch where I can compile with cst but not inlining? I need to debug %defcallback because just as I suspected - when I try to compile cando with the new dev it fails in the worst place possible. While compiling the pzmq system. Source tracking could help me debug it.
5:53:41
drmeister
::notify Bike What happened to compile-lambda-form-to-llvm-function? I need it to make defcallback work - can revivify it https://github.com/clasp-developers/clasp/blob/mps/src/lisp/kernel/cleavir/translate.lisp#L1176
10:05:05
Colleen
attila_lendvai: drmeister said at 2018.06.12 21:30:28: I ran into the problem you saw on linux "error: can't create dynamic relocation R_X86_64_PC32 against symbol:" - I'm trying to figure out a work around. Can stuff just keep working?
10:05:05
Colleen
attila_lendvai: drmeister said at 2018.06.12 21:57:09: It looks like a workaround is to not use the lld linker when building in object mode on linux (phew). I added support for lld for bitcode/lto mode - so it's no big deal to switch back to ld for object mode.
10:05:06
Colleen
attila_lendvai: drmeister said 21 hours, 54 minutes ago: I made a change to the parallel scraping code - so that it always does 8 parallel scraping jobs. It looks like every time we pass a different -j xx argument to ./waf it starts scraping everything all over again. My compromise hack is to stop that from happening. We need a better way
10:39:58
attila_lendvai
::notify drmeister hrm, I could reproduce the waf jobs issue locally. a note: it's better to record the reason for the change right into the wscript file in a comment. I've seen that diff long ago, versus just saw this ephemeral note here on irc. code comments also remain near the change for long if someone later on wonders about it. I've pushed a commit to record it there.
10:43:11
attila_lendvai
::notify drmeister re that linker issue: reading the bug reports it look to me a legit error that happens due to not doing something properly, and it may not cause any issues when linking in a more permissive environment (macos?). but it's just a vague impression...
10:47:15
Colleen
drmeister: attila_lendvai said 7 minutes, 17 seconds ago: hrm, I could reproduce the waf jobs issue locally. a note: it's better to record the reason for the change right into the wscript file in a comment. I've seen that diff long ago, versus just saw this ephemeral note here on irc. code comments also remain near the change for long if someone later on wonders about it. I've pushed a commit to record it there.
10:47:15
Colleen
drmeister: attila_lendvai said 4 minutes, 4 seconds ago: re that linker issue: reading the bug reports it look to me a legit error that happens due to not doing something properly, and it may not cause any issues when linking in a more permissive environment (macos?). but it's just a vague impression...