@beagle3 right, this must be my fault

@JeromeIbanes it should be. isn't it?

@ngn it is, that's great.

@JeromeIbanes t (the test runner) starts k as an external process, gives it input, and collects its output

yep

are you comparing the output to the expected result?

@JeromeIbanes yes

very cool

700 external processes and it takes only half a second :)

(that's because i don't wait() for them to finish but don't tell anyone)

on aiohttp, I commonly run at 1000 queries per second on my old i5 (per process)

nice

@JeromeIbanes i completed 100 make-s without any mysterious failures - hopefully they're all gone now

@ngn great, thank you!

yeah it works!

thank you

@JeromeIbanes can you give me a sample of the data you'd typically want to serialize? it would be nice to have a benchmark to measure the impact of validation

json data, k expressions, dictionaries, I can email you a sample if you want?

@JeromeIbanes yes please

will do

sent you a sample

@JeromeIbanes thanks!

for a benchmark, representative is good enough

@JeromeIbanes is each line meant to be serialized individually or the whole file as a single k table?

each line, as a matter of fact, the number of keys is variable per line

ah, right, it's not uniform

can you repro the issue?

@JeromeIbanes which one?

the one in the mail

@JeromeIbanes no, it works for me

interesting

@JeromeIbanes good news, after running it 20 times it crashed

btw, you can write 0:"a.txt" instead of 0:`$":a.txt"

oh thanks, I guess bad habits haha

how big was the core file?

@JeromeIbanes core file? i don't know how to generate or use those

unlimit -c unlimited

but no worries, I'm surprised it didn't make one by default

cat /proc/sys/kernel/core_pattern

will tell you which directory/file the core is saved to

you can/could change this value

and then what? i'd have a large dump of ram content without the slightest idea of where to look for the error :)

hehe true