hi, unix question if you don't mind. when using fread, it's possible to detect error from EOF using the feof and/or ferror interfaces, however, can such read error be detected in a shell script reading from stdin? can stdin have read error at all? if yes, as far as i know "read" does not distinguish read error from EOF, but data=$(cat) (with the caveats of LF truncation and finite stream) might error on read, right?

the goal is to be absolutely sure that as far as the script can tell, the whole input has reached the shell script successfully.

avih: If there a short read of the file then cat would return an error and exit non-zero.

w

bahamat: thanks. i think the first question is whether at all stdin can have read error which is distinguishable from EOF. first in a C program, and then in a shell script. and if the answer is yes for a shell script, then continue to ways of detecting this case

avih: Well fread() and a shell script using cat are different things. A userland process using fread is supposed to use ferror/feof. If it doesn't, that's probably a bug in that program. If you're writing a shell script you don't need to worry about what functions or system calls are being used. However the command behaves is the interface you need to deal with.

ok, let's forget the shell script. if a c program uses fread on stdin, is there ever a use case where ferror returns true?

if yes, how to synthesize it?

i.e. how to invoke/set-it-up so that ferror(stdin) ends up true?

it's easy to synthesize a use case where fwrite ends in ferror, e.g. if the receiving side of pipe closes, then the writing side gets a write error. but how is it possible to get a read error? in a pipe, if the writing side closes, then i don't think the reading side gets an error. it only sees feof, not ferror