[Toybox] (no subject)

[Rob Landley]

Catching on on email on the plane back to Austin...

On 09/15/2013 04:23:09 AM, hhm wrote:
> Found 2 bugs in `od`, not very familiar with the code so just  
> reporting them:
> 
> 
> 1. -t <integer-type>0 does not work correctly (loops infinitely while
> printing 0)

Confirmed. (Although "correctly" in this case is producing an error  
mesage.)

And the help text for this command is horrible. And it's accepting -c  
but not checking for it...

Hmmm, the help text mechanism isn't flexible enough to chop out mention  
of floating point from the help text when the config has disabled  
floating point. Yet another todo item, but I have no idea how to  
address that one. (Apply the C preprocessor to the file before chopping  
out help text? I guess comments aren't touched by the preprocessor.  
Still, that'd slow the build down even more...)

Worry about it later...

> 2. -J does not seem to have an effect (although it looks like there is
> some code which does reference/use it)

Yeah, that's not working either. Hmmm...

For starters, the jump_bytes < pos test is backwards, it needs to do  
work when jump_bytes is > pos. Secondly, lskip() returns the amount of  
data _left_, not the amount consumed. And the implementation of lskip()  
is wrong. (The point is to make skipping -J through multiple files work  
like you catted files into stdin, so it has to accurately track how  
much it consumed from each file...)

Hmmm, tension between od currently being the only caller of lskip() and  
wanting to make lskip() generic (not assuming it's at the start of the  
file; in theory it doesn't need to check the current position, it  
should always be zero, but I don't want to assume it'll never be used  
elsewhere...)

What lskip() does is lseek() that falls back to read() in a loop to  
advance if lseek doesn't work (ala stdin when it's a pipe), but toybuf  
is already used and libbuf should only be used from library functions  
and I dowanna declare a 4k local variable on the stack. Besides, mke2fs  
would presumably use this... except on output, not input. Sigh.

Ok, fixed it up for now. I should add test suite entries for this at  
some point. (Upload to site happens when I get my netbook and laptop  
talking to each other again, probably tomorrow.)

Thanks,

Rob