Eric Blake <ebl...@redhat.com> writes: > On 05/15/2018 10:26 AM, Andrew Jones wrote: >> On Tue, May 15, 2018 at 04:48:33PM +0200, Igor Mammedov wrote: >>> When using following CLI: >>> -numa dist,src=128,dst=1,val=20 >>> user gets a rather confusing error message: >>> "Invalid node 128, max possible could be 128" >>> >>> Where 128 is number of nodes that QEMU supports (MAX_NODES), >>> while src/dst is an index up to that limit, so it should be >>> MAX_NODES - 1 in error message. >>> Make error message to explicitly state valid range for node >>> index to be more clear. >>> >>> Signed-off-by: Igor Mammedov <imamm...@redhat.com> >>> Reviewed-by: Andrew Jones <drjo...@redhat.com> >>> --- > >>> if (src >= MAX_NODES || dst >= MAX_NODES) { >>> error_setg(errp, >>> - "Invalid node %d, max possible could be %d", >>> - MAX(src, dst), MAX_NODES); >>> + "Invalid node %d, The valid node range is [0 - %d]", >> ^ should be a '.' >> >> And maybe need a '.' at the end of the second sentence too, as it's not >> an error phrase, but a real sentence. >> >>> + MAX(src, dst), MAX_NODES - 1); >>> return; >>> } > > Actually, error_setg() is documented as taking a single phrase (no '.' > included), and that if you need a second sentence, it's better to use > error_append_hint().
Correct. Providing help on valid values is exactly what error_append_hint() is for. > Maybe Markus has an opinion on the best way to > word this error message. Yes: "Parameter 'src' expects an integer between 0 and 127" Referring to an erroneous key=value by value is not nice. What if the value occurs in multiple places, and is valid in at least one? key is there, it's unique[*], so use it. [*] Except in the few places that use repeated keys to form lists. Ugh.