Search
Monday, 23rd of October 2017, 10:26:20 UTC
15:21:19
k-os
i'm trying to compile on arm again, and again iolib is breaking
15:21:19
k-os
this is the error message coming out of iolib/src/syscalls/pkgdcl.lisp
15:21:19
k-os
Condition of type: SIMPLE-PACKAGE-ERROR
15:21:19
k-os
There exists no package with name "IOLIB/COMMON-LISP-USER"
15:23:10
k-os
some clarifications: it's compiled to an executable, compilation was successful, executing not, on x86_64 to x86_64, it should be loading iolib.base first
15:23:44
jackdaniel
k-os: did you try with ECL from the repository?
15:23:57
k-os
yes it's the develop branch
15:24:06
k-os
with the most recent asdf 3.3.01
15:24:18
jackdaniel
ASDF had a bug with systems named foo/bar (with slash), but it should be solved. other than that, I don't know what may cause the error
15:24:39
jackdaniel
so it does work on x86-64 or it doesn't?
15:24:41
k-os
yes i could solve that bug by renaming ^^
15:24:52
k-os
no it doesn't work on x86_64
15:25:40
k-os
minimal use case with only iolib as a dependency and it still breaks
15:26:04
jackdaniel
ah, so it is about package name not a symbol name?
15:26:12
jackdaniel
you have changed package name and it started working?
15:26:49
k-os
that's how I solved it the last time
15:27:33
k-os
the problem seems to be the order of loading
15:27:46
k-os
it should load iolib.base first as a dependency of iolib.syscalls
15:27:57
k-os
but it starts with the files of iolib.syscalls
15:31:04
k-os
should I open an issue on gitlab?
15:31:17
jackdaniel
if I got a penny for each time iolib breaks on ECL I would have a dozen of them ;)
15:31:24
jackdaniel
yes, please do, with steps to reproduce
15:32:11
k-os
but it's not only iolib, quri, cffi, clsql are broken for ecl binary compilation
15:33:20
k-os
but they are easier to solve, most are wrong dependency declarations in asd files, and some load files at run-/compile-time
15:37:15
jackdaniel
if it is a bug in a library, then such bug should be reported in a library
15:38:09
jackdaniel
I will take a look if you report the issue
15:38:17
jackdaniel
right now I have to go, see you o/
15:42:43
pjb
It's actually a little strange, since one would think ecl is the easiest implementation to do posix stuff from.
15:43:29
jackdaniel
the thing is, that iolib is developed on sbcl
15:43:37
jackdaniel
so it is ECL doeing posix stuff sbcl-way
15:43:48
jackdaniel
k-os: could you also try, if this breaks with the newest ASDF release?
15:45:38
jackdaniel
k-os: when I run ecl -norc -load ~/.quicklisp/setup.lisp -load base.asd on the file you have pasted nothing of this kind happens
15:45:43
jackdaniel
should I type sometihng in the repl?
15:46:22
jackdaniel
(I have pointed to ~/quicklisp/setup.lisp in my case, I have default directory)
15:50:14
k-os
ah did you (asdf:make :base)
15:50:49
k-os
my bad i omitted some steps
15:50:54
jackdaniel
I'm going to a gym, see you o/ please fill information about results with the upstream ASDF and add this command (asdf:make :base) to the description.
15:51:49
jackdaniel
(how to test with upstream asdf: ecl -norc -load /path/to/upstream/asdf.lisp -load ~/quicklisp/steup.lisp -load base.asd -eval '(asdf:make :base)'
16:00:45
k-os
yeah for some reason I forgot to git pull on asdf, but nothing changed with upstream
16:05:40
k-os
maybe important it's iolib v8.3
16:14:37
jackdaniel
I can't reproduce it on x86-64 with ECL 16.1.3 nor with ECL from the develop branch
17:09:38
k-os
jackdaniel: just to check, you run the "base" executable after you leave the repl
17:10:01
k-os
also i replaced iolib with osicat in the asd description and it also fails but missing package is alexandria
17:18:00
k-os
also replacing iolib with alexandria and using and not using its functions works fine
18:10:47
k-os
i have an even simpler example, two system declarations and two files
19:03:14
jackdaniel
did you try with make-build ?
19:03:49
jackdaniel
arguably this is asdf bug, but I'll try to reproduce it at some point of time
19:03:55
jackdaniel
I'm exhausted right now :)
19:05:52
jackdaniel
this is ECL-specific interface, documented in ECL documentation
19:11:06
k-os
i'm not so sure if the simpler version exhibits the same problem, the build-path probably overwrites the package fasl and that was the problem of the simpler version
19:14:32
k-os
same error with make-build
19:15:11
jackdaniel
I will try to reproduce it on arm, because I don't see what you see on x86-64
19:15:21
jackdaniel
did you try to clean up ~/.cache/common-lisp ?
19:16:27
jackdaniel
no, but I will update somewhere this week and see
19:17:38
k-os
i tried to compile with fasb
19:18:34
jackdaniel
from ecl you may do (load "path/to/file.fasb")
19:18:42
jackdaniel
and you have all systems loaded
19:20:02
k-os
maybe you could use the fasb? or is it system-dependen like sbcl fasl
19:21:32
jackdaniel
it might have worked given we'd have same configure options etc etc, but it rather won't
19:21:41
jackdaniel
either way I won't have time to work on it this week
19:21:57
jackdaniel
I'm full, maybe I'll take a look over the weekend and try to reproduce with 7.2
19:22:52
k-os
which gcc version are you using?
20:29:01
k-os
gcc 6.4.0 (x86_64), 4.9 (arm 32) same error
Monday, 23rd of October 2017, 22:26:20 UTC