11:13 <discocaml_> <pigeon.v> `Uri.to_string` uses comma-separated values for non-singletons:

11:13 <discocaml_> <pigeon.v> ```

11:13 <discocaml_> <pigeon.v> # Uri.(to_string (with_query (of_string "http://example.com/x") ["a",["b";"c"]]));;

11:13 <discocaml_> <pigeon.v> - : string = "http://example.com/x?a=b,c"

11:13 <discocaml_> <pigeon.v> ```

11:19 <companion_cube> Can't wait for Uri to not depend on cstruct or whatever...

13:01 <Anarchos> i have a strange bug with the test lib-format/mc_pr586_par2.ml

13:01 <Anarchos> sometimes it prints 42f+01 isntead of 42.000000

13:02 <Anarchos> i don't know how to investigate but hte test mc_pr586_par which does not use Format.synchronized_formatter_of_out_channel works perfectly

13:07 <discocaml_> <dinosaure> You should make an issue 🙂

13:14 <companion_cube> He's on an unsupported platform

13:53 <discocaml_> <leviroth> That was a reply to your message about uri/cstruct

13:57 <octachron> Anarchos, you should try to reproduce the issue with only call to fprintf "%f" on different formatter C in parallel, in the issue is still present, try to test the Haiku implementation of vsnprintf in C only in parallel threads

14:06 <Anarchos> octachron on different formatter C in parallel ? What do you mean ?

14:07 <octachron> on different formatters in parallel.

14:08 <octachron> to rule possible issues with domain-local storages or synchronized formatters.

14:08 <octachron> *rule out

14:08 <Anarchos> octachron ok

14:12 <Anarchos> octachron mc_pr586_par.ml works but mc_pr586_par2.ml exhibits the trouble

15:01 <discocaml_> <darrenldl> anyone knows how to specify a windows path in eio?

15:02 <discocaml_> <darrenldl> i've tried `fs / "C:/file"`, `fs / "C:\file`, `fs / "C/file`, `fs / "/C/file"`, and none of them worked

15:39 <octachron> Anarchos, there are not much difference between the two tests, you could try to inline the definitions, but it probably better to increase the test duration to see if this is not an accident.

17:20 <Anarchos> octachron it is a random behaviour. some times the test succeeds some times it fails

17:22 <octachron> Yes, this is clearly a data race issue, the aim is to discover where is the data race.

18:30 <discocaml_> <otini_> what’s your architecture?

18:31 <discocaml_> <otini_> if ThreadSanitizer can spot the data race, that would simplify things

18:32 <discocaml_> <octachron> The problem is probably more the OS (Haiku) than the architecture.

18:32 <discocaml_> <otini_> a data race inside the OS?

18:35 <discocaml_> <octachron> I meant in term of getting TheadSanitizer to work. And if I am not mistaken, Haiku has its own C standard library.

18:36 <discocaml_> <otini_> Oh, ok, I get it

19:45 <discocaml_> <limp.biskit> never see haiku actually used in the wild

19:45 <discocaml_> <limp.biskit> cool project

22:22 <Anarchos> limp.biskit yes it is