freenode/#clasp - IRC Chatlog
Search
13:39:00
drmeister
Bike: it looks like poiu might be a thing. Can you pull it and run the tests in sbcl to see how it works there?
13:40:17
drmeister
I added the support for it to clasp but it doesn’t compile things and I don’t have time to debug Asdf
13:42:54
drmeister
Asdf and poiu try to load system files when it should be detecting that they aren’t there and compiling them and then loading them.
13:44:21
drmeister
This may be a bug in clasp not recognizing that they are not present or asdf uses an error signal from load to detect that the file isn’t available and we are signaling the wrong error
13:46:20
drmeister
I say that because if I continue from the error it compiles the file and then fails on a subsequent load of a file that isn’t there.
13:47:31
drmeister
I have no evidence that asdf/poiu are using fork or waitpid or alll that good stuff that would speed up and parallelize building.
13:48:51
drmeister
The example that fare provides for point testing is a complicated one that messes with readtables and it reveals another problem.
13:49:28
drmeister
Our eclector.readtables:readtablep doesn’t have a default method that returns nil.
13:52:08
drmeister
Continuing: so readtablep fails if the arg is not a readtable rather than returning nil
16:33:27
drmeister
I did a quick survey of systems that provide :depends-on information that might be good tests for poiu
16:35:13
drmeister
stassats: I'm not certain yet - I'm trying to figure that out. fare posted an update to the README.md that suggests that it does.
16:36:08
drmeister
"POIU 1.34 reportedly passed its test of building and using Exscribe with ASDF 3.3.2.5 on SBCL 1.4.4 on Linux x86-64."
16:36:18
drmeister
"Still, support for ASDF 3.3 is considered incomplete. See TODO section at the end."
16:36:45
drmeister
I'm desperate enough to try. But I haven't done the obvious thing of testing it on sbcl.
16:39:17
drmeister
There - I just tested it on sbcl and it fails in the same way that it fails on clasp.
16:40:05
drmeister
But now I've added support for clasp to poiu and vice versa and I'm more invested in getting it to work.
22:01:19
kpoeck
drmeister to be able to (ql:quickload :exscribe) you need to load the first code block from https://github.com/clasp-developers/clasp/wiki/Properly-integrate-eclector-into-clasp
22:03:06
kpoeck
Caveat, clasp complains twice that a dispatch-macro-character is being redefined (CORE:SHARP-DOT and CORE:SHARP-LEFT-PARENTHESIS)
22:05:26
drmeister
Thank you - we should integrate much or all of that code block into the clasp source code.
22:06:15
drmeister
But it looks like poiu may be broken - I'm not sure about that. I've asked cracauer to investigate because I'm getting tied up with company stuff again.
22:08:03
kpoeck
right now we only use eclector to compile files and not to load (or simply read), have a proposal for that
22:09:06
kpoeck
crosscompiling sbcl is the best test, they do funny things to the readtable in "make-host-1".
22:09:43
kpoeck
sbcl also use structs to an extreme, another great test for our defstruct integration
22:10:34
drmeister
I made one followup change to that PR you submitted. I moved the (translate-logical-pathname pn) into write-faso.
22:11:13
drmeister
I think I should do that as a rule for any function that reads from a file specified by a file-name - run it through translate-logical-pathname
22:12:25
drmeister
No worries. You can go on (core:write-faso<point> and hit M-. and it will jump you to the C++ code. It is a bit of a wilderness in there though. Take a look at what I did - it should be perfectly understandable to you.
22:18:37
kpoeck
A chance that anybody for your group come to els in Zurich? For me its just a 4 hour train ride, so i will come
22:45:28
selwyn
i'm going to try and implement everything on my own and see how it goes, so far its going well