Re: documentation bug re character range expressions

[Marcel (Felix) Giannelia]

On 2011-06-03 05:09, Greg Wooledge wrote:
> Oh, look, there's more!
>
> [...]
>
> See?  Both tr(1) and ls(1) do it too!

Right; forgot about ls (because "alias ls='LC_COLLATE=C ls'" has been in 
my .bashrc for so long that I completely forgot it was there :) ), and 
didn't think to try tr -- but tr appears to be case-sensitive under all 
locales for me, though.

And yours looks broken -- how does
echo Hello World | tr A-Z a-z
result in a bunch of non-ASCII characters? (That's how it looked when it 
got here, at any rate -- maybe one of our mail servers did something.)

> > Even grep, whose man page says it obeys LC_COLLATE and the locale,
> > actually has [a-c] equivalent to [abc] on all locales. Someone must have
> > snuck in and fixed it.
>
> You must live in a strange and peculiar world.
>
> imadev:~/qwerty$ type grep
> grep is hashed (/usr/bin/grep)
> imadev:~/qwerty$ echo 'brown cow' | grep '[A-C]'
> brown cow
> imadev:~/qwerty$ echo 'BROWN COW' | grep '[a-c]'
> BROWN COW
>
> Is every single bit of your knowledge born out of familiarity with just
> ONE operating system with weird extensions?

I tried this on a few systems with different distros on them. The only 
non-Linux systems I have access to are running things so old that they 
only have the C locale, so I can't be sure about them. (For instance, 
the only Solaris system I can get at doesn't support unicode at all.)

Here are a couple:

[c69:~]$ uname -sr
Linux 2.6.32-31-generic
[c69:~]$ type grep
grep is /bin/grep
[c69:~]$ locale | grep LC_COLLATE
LC_COLLATE="en_US.UTF-8"
[c69:~]$ echo 'brown cow' | grep '[A-C]'
[c69:~]$ echo 'BROWN COW' | grep '[a-c]'
[c69:~]$


u-elive ~(0)$ uname -sr
Linux 2.6.32-26-generic
u-elive ~(0)$ locale | grep LC_COL
LC_COLLATE="en_US.UTF-8"
u-elive ~(0)$ type grep
grep is /bin/grep
u-elive ~(0)$ echo 'brown cow' | grep '[A-C]'
u-elive ~(0)$ echo 'BROWN COW' | grep '[a-c]'
u-elive ~(0)$


The university used to have a big AIX system, but again, I think it was 
a version from before the days of unicode locales. Maybe grep working as 
expected is a Linux thing?

And if so, then you're right about me living in a "strange and peculiar 
world" -- it's the world of people and organizations too poor to afford 
proprietary Unices; one would never see HP-UX in a world like that when 
there are cheaper alternatives. I'll spin up some BSD's in a virtual 
machine and check those later -- anything else I should try?

> You ought to report the bug in your vendor's grep(1) implementation, if
> it is actually broken as you describe.

And no, I'm going to keep very quiet about this "bug" in grep -- because 
it's working the way I want/expect it to, and it'll doubtless break 
many, many shell scripts and cause data loss for a fair number of people 
if it were fixed :)

~Felix.