Re: Possible bug to report (loss of data)

[Harry Putnam <reader@newsguy.com>]

Tony Theodore <tonyt@logyst.com> writes:

> Maybe something other than the output of locale:
>
> $ locale
> LANG="en_AU.UTF-8"
> LC_COLLATE="en_AU.UTF-8"
> LC_CTYPE="en_AU.UTF-8"
> LC_MESSAGES="en_AU.UTF-8"
> LC_MONETARY="en_AU.UTF-8"
> LC_NUMERIC="en_AU.UTF-8"
> LC_TIME="en_AU.UTF-8"
> LC_ALL=

Tony wrote:
>> I suspect this error is an environment variable issue - check the
>> output of "locale" and maybe try something like:

What made you suspect Locale ENV problem?  Just a guess or something
that seems to indicate that kind of trouble stood out to you?

I've had locale ENV trouble once when I installed opensolaris.  The
manpages all had some amount of gobbledeegook in them.  As I recall it
required some kind of env setting but I've forgotten exactly what it
was but involved setting things to something other than utf8.

When I try an experiment that recreates the moves made when the
problem occured, like this:

    tarsnap -c -f my_etc etc
  
  Followed by the copy to a new name:
   tarsnap -c -f my_etc-$(dtf)  @@my_etc
  
  Then followed with delete of the original:
   tarsnap -d -f my_etc
  
I see no problems whatsoever.  
Checking the sizes with --print-stats on the original and the copy
shows no discrepancy at all.  

The error never comes up.  But then I didn't enclude my blunder as
posted.  That is trying to copy by having the names backward.

  -f oldname @@newname

Instead of 
  -f newname @@oldname (the correct way)


To recreate the actual setup the way it was when the error occured
would require an upload of some 44 gb.  (I wonder if that long of read
when coping could be part of the problem...) But since I need to do it
anyway (since my archive has gotten slaughtered).. I'll report back.
It takes about 2 whole days for the upload.

However it seems unlikely that the problem would only occur on larger
archives if its and ENV related problem eh?  That is, unless the
lengthy read is somehow involved.

Also seems unlikely that I'd need any special ENV setting since the
experiment worked fine without that.

-------        ---------       ---=---       ---------      -------- 

For me locale shows things nearly all set to `POSIX':

  LANG=
  LC_CTYPE="POSIX"
  LC_NUMERIC="POSIX"
  LC_TIME="POSIX"
  LC_COLLATE="POSIX"
  LC_MONETARY="POSIX"
  LC_MESSAGES="POSIX"
  LC_PAPER="POSIX"
  LC_NAME="POSIX"
  LC_ADDRESS="POSIX"
  LC_TELEPHONE="POSIX"
  LC_MEASUREMENT="POSIX"
  LC_IDENTIFICATION="POSIX"
  LC_ALL=