12:18 lizmat left, lizmat joined 14:50 Geth left 16:27 Geth joined
[Coke] "rak thing dir" - no match. "rak thing dir/file" -path specification did not select any files. - but the file exists, and has this string in it. 17:12
is there a way to generate debug output so I can see why it's not finding it? 17:13
lizmat --rak 17:55
[Coke] Path specification did not select any files, paths: [".\\config\\my-file.json"], dir .\config\my-file.json - it's there. 18:03
might it think it's not a text file? 18:04
from the readme: just before it isi being called. 18:06
small typo
if I explicitly add "--extensions=json", it finds it. 18:07
lizmat is it a very small file ?
also: is it readable? 18:08
[Coke] 2960L, 103967B
"type path/to/file" works. 18:09
I can vi it
(this is windows)
rak can read it fine if I use --extensions=json
I don't have a .rak-config.json in my home dir on this windows box 18:13
lizmat hmmm... wonder if it is somehow Windows related...
[Coke] any reason why the first run of 'rak thing dir' might be significantly slower than future calls? 18:15
Guessing it has to be the is-text= true default 18:16
yah, specifying "extensions" doesn't show is-text then in --rak 18:17
I also do not see that I opened this as an issue previously. You're using a module for the "is-text" check, I'm guessing? 18:18
lizmat what does: use path-utils; say path-is-text("config/my-file.json")
say?
[Coke] 0 18:19
lizmat ok, so it considers it to be binary
assuming the path is correct 18:20
maybe the file has a BOM? and that triggers it ?
also the test assumes it's UTF-8 or ASCII
any 16bit encoding would probably also consider it binary 18:21
[Coke] vi -b shows no BOM
it does show a ^M per line...
I'll figure out what in the algo is causing the fail 18:22
lizmat it *is* open source... :-) a PR for path-utils would be brill 18:26
[Coke] Having this all be in NQP makes debugging hard. 18:29
Oh. looking at just comments: 18:36
# \r not followed by \n hints binary
... that is not true for windows
lizmat no?
[Coke] ... ah, whoops, it's expected to be \r\n windows, \n linux. oops 18:38
lizmat yeah, old pre-OSX macs only had \r only
[Coke] got it. You are trimming the buffer *first*, then checking to see if the \r is followed by an \n. 18:43
... and in this file, guess what the buffer ends on.
Just happens to be split at the magic 4096 mark. 18:44
if I change it to 4095, it says it's text 18:45
or, more importantly, 4097.
opened a ticket. 18:47
path-utils #3
lizmat ah, cool. thanks!
[Coke] \o/ 18:48
lizmat line 88 in path-utils: 18:55
if you change that to: my int $limit = nqp::elems($buf) -1;
does it accept it then? (I think it should)
[Coke] af(that)k right now, will check later and report back 19:18