The man page for binary should mention that NaN and Inf (and any other substituted values, even if they are being substituted correctly and by the OS), so that the developer is aware of the potential of the data .

I think it is our intention that Inf and NaN should just be handled correctly here, Bug [85ce4bf928] notwithstanding. If the bit-pattern specifies those values, they should be scanned correctly. If the arguments specify those values, they should be formatted correctly. (This is unlike the handling of NaN in expr.)

I'm not aware of the original context of this ticket, but it may relate to [85ce4bf928], which showed that [binary format] will not store Inf.  The original discussion (if it can be unearthed) might be helpful to understand background for [85ce4bf928].

The suggestion was that for the Tcl binary.n man page, if it is discovered that some implementations replace input values with NaN, Inf, or some other string , a note should be added to the man page so a developer, trying to discover why their Tcl program is behaving in an unexpected manner, could see the potential issue.

I believe this was opened after a discussion occurred indicating that use of binary might result in Inf or NaN instead of the binary value expected.

Incomplete request?

Logged In: YES 
user_id=80530
Originator: NO


I confess I don't understand what
the problem is, so I don't know
how best to fix it.