ryanc
2017-2-27 14:41:29

@mflatt a fresh checkout of mflatt/racket#linklet segfaults during make for me (building on Linux) with SIGSEGV MAPERR si_code 1 fault on addr 0x39. It appears to be somewhere in the db package when the error occurs.


mflatt
2017-2-27 15:25:00

@ryanc I’ll push a repair soon; I’m not sure why it doesn’t crash on Mac OS, but the reason for the crash seems clear, and it’s related to yesterday’s sync with the main branch


ryanc
2017-2-27 15:28:38

@mflatt thanks!


spall
2017-2-27 16:20:41
echo '(reset-handler abort) (keyboard-interrupt-handler abort)' '(compile-file "core.sls")' \| scheme -q
compiling core.sls with output to core.so
Exception: missing import for make-mutex (only (chezscheme) make-mutex) at line 1, char 9 of <http://lock-hash.ss\|lock-hash.ss>
make: *** [core.so] Error 255
```   the offending line
```(import (only (chezscheme) make-mutex))``` Maybe i am not allowed to put an import in this file?  I'm having a hard time figuring out how to correctly use chez.
<strong>@mflatt</strong> <strong>@samth</strong>

```spall@spall:~/iu/hash-tables/not-a-box$ make
echo '(reset-handler abort) (keyboard-interrupt-handler abort)' '(compile-file "core.sls")' \| scheme -q
compiling core.sls with output to core.so
Exception: attempt to reference unbound identifier make-mutex at line 5, char 12 of <http://lock-hash.ss\|lock-hash.ss>
make: *** [core.so] Error 255

Without the import I get this error.


samth
2017-2-27 16:21:31

@spall maybe come by my office if you’re in LH


spall
2017-2-27 16:21:35

alright


mflatt
2017-2-27 16:26:02

@ryanc Should build ok, now


ryanc
2017-2-27 16:48:10

@mflatt yes, it works for me now, thanks


samth
2017-2-27 20:04:35

I consistently get this error when running make in not-a-box:


samth
2017-2-27 20:04:43
/home/samth/tmp/linklet/racket/collects/racket/private/kw.rkt:773:47: read-syntax: bad syntax `#λ`
  context...:
   expander.scm:59938:18
   expander.scm:60759:51
   expander.scm:61074:65
   expander.scm:61099:65
   expander.scm:61099:65
   expander.scm:61099:65
   expander.scm:61099:65
   expander.scm:61099:65
   expander.scm:61099:65
   expander.scm:61099:65

samth
2017-2-27 20:04:57

this is with HEAD of both repos


mflatt
2017-2-27 20:09:22

I’m not getting that at the moment, and I don’t have a clear idea of why the reader would receive a λ in place of a t


samth
2017-2-27 20:10:06

alternatively when I used a different install of linklet I got a paren-vs-square bracket reader error


samth
2017-2-27 20:10:17

I’ll try a fresh clone of everything


mflatt
2017-2-27 20:10:31

It could be something in the port layer that depends on how many bytes an input port produces at a time, but it seems strange that you get consistent errors and I don’t get any


samth
2017-2-27 20:10:56

could it have to do with on-demand-zo-loading?


samth
2017-2-27 20:11:17

sadly that didn’t help :disappointed:


mflatt
2017-2-27 20:11:27

That did seem unlikely


mflatt
2017-2-27 20:11:47

You’re using a stock Chez?


samth
2017-2-27 20:12:28

well, with my PR about date-dst?


samth
2017-2-27 20:12:32

but this happened before then


samth
2017-2-27 20:12:37

I’ll try bisecting


samth
2017-2-27 20:12:43

once these tests finish


mflatt
2017-2-27 22:19:14

@samth I get that error when I try to build on Linux, so I can investigate more


samth
2017-2-27 22:19:27

ok, great


samth
2017-2-27 22:19:37

my guess is that it’s an encoding issue


samth
2017-2-27 22:58:09

Inconveniently, I can’t find any revision on which I can successfully read that file


samth
2017-2-27 23:16:07

definitely an encoding issue, since commenting out the entire file with ; causes the reader to loop


mflatt
2017-2-27 23:25:21

It really was something in the port layer that depends on how many bytes an input port produces at a time


samth
2017-2-27 23:28:34

fun


samth
2017-2-27 23:33:52

works great now, thanks!


samth
2017-2-28 02:01:44

Looks like the changes have fixed the performance problem for just running racket code — my little bubble sort benchmark now runs as fast as extracted (and 30% faster than regular racket)


mflatt
2017-2-28 03:21:03

Good – I was hoping for that. I’ve been looking at the regexp expander as an example, where changes haven’t made much improvement, yet