asciilifeform: i suspect an oddity with the linux tty subsystem. the Right Thing solution would be to find it. the alternative, ugly solution, is 1) 512 byte buffered reads 2) with timeout .☟︎
asciilifeform: diana_coman: ok, finally managed to get a stuck dd
asciilifeform: mod6: recall, that one he asked to let be until expires
asciilifeform: ( how exactly this ties the hands of, e.g. , taiwanese fabs, i do not presume to know , but result is evident )
asciilifeform: they're actually named by name on usg 'export ban' list.
asciilifeform: i've turned up some hints that usg actually interferes with the ADC makers -- 'ohnoez, moscow could use fast ADC in radar set'
asciilifeform: i.e. are pressed shitboard , rather than hardwood.
asciilifeform: and they ~all do shenanigans with multiple slow ADCs masquerading as 1 fast one
asciilifeform: doubtful, all of the examples i've been able to peek into, are full of xilinx.
asciilifeform: ( in this context, 'sane' ~= sits on pcie bus, and ADC turnaround is fixed # of bus cycles, preferably 1. and documented i/o. )
asciilifeform wouldn't consider , say, 1btc , for a ~sane~ scope, to be unreasonable
asciilifeform: so if you want a 1G/s scope with os where you actually have some control of the timing ( for e.g. real time phase sampling ) you're 100% phuqed.
asciilifeform: ( these, for extra laughter, come with ~built in windows~ in the instrument itself ! )
asciilifeform: mircea_popescu: note, not in the i-dun-give-half-a-fuck-what-it-costs market, either
asciilifeform: sorta hilarious , this, 'here's pc with very spiffy 6G/s pcie bus, for 100$, but if you want to connect something useful to that, haha, what do you think this is, the fyootoor'
asciilifeform: not for 100 $, not for 100, not for 10,000.
asciilifeform: in other noose, since 1st time i asked the q in 2013, appears that there is ~to this day~ not such a thing as a 1Ghz oscilloscope with any linux support whatsoever.
asciilifeform: good % of unixland, by weight, consists of this.
asciilifeform: 'if i actually fix it, would have to admit that it was buggy'
asciilifeform: recall , similarly, koch's 'fix' for his mpi bug.
asciilifeform: afaik to this day no such provision in any unixlike.
asciilifeform: ( naturally the imbecile kernel doesn't actually test whether something is riding on the tty prior to opening, 'why would we' )
asciilifeform: mircea_popescu: happened to me '9000' times when testing FG, and earlier iron.
asciilifeform: btw if anyone wants to experimentally observe the zombie effect, can do it with ordinary gnu 'screen' -- when opened on a tty with blocking i/o, will routinely leave zombie turd instead of properly terminating, then hours later you find that tty is wedged, and discover 9 copies of 'screen' in process table.
asciilifeform: 'and for extra sport, i will also shit where you cannot clean'
asciilifeform: and ought to ring all of the alarm bells.
asciilifeform: a FG that produces , under normal conditions, 0 output for, say, 30 seconds, is indicative of dead iron.
asciilifeform: this being said, if you implement a timeout for the blocking reads, you will not have zombie problem.
asciilifeform: ( asciilifeform did not write the linux i/o subsystem ; nor the pl2303 kernel mod ; and will readily admit that they are sad. and if anyone has idea for some other means to connect a 115200 baud box to machine, that does not suffer from the weakness of these, i'm as always all ears )
asciilifeform: well not each time, but to see if the 'init breaks the dongle' hypothesis is troo.
