Bash bug?

[Robert Nichols]

Zoltan Boszormenyi wrote:
> Zoltan Boszormenyi írta:
> 
>> Hi,
>>
>> I am trying something like the following,
>> with a configuration file containing a token and
>> a directory in a line, depending on the tokens,
>> certain actions should be taken later. Validating
>> the configuration file whether all the required/optional
>> tokens are in the file should go like this:
>>
>> -----a.txt----------------------
>> A directory1
>> B directory2
>> C directory3
>> --------------------------------
>>
>> -----a.sh-----------------------
>> #!/bin/bash
>>
>> HAS_A=0
>> HAS_B=0
>> HAS_C=0
>> cat a.txt | while read i ; do
>>         if [ "`echo $i | awk '{ print $1 }'`" = "A" ]; then
>>                 HAS_A=1
>>         fi
>>         if [ "`echo $i | awk '{ print $1 }'`" = "B" ]; then
>>                 HAS_B=1
>>         fi
>>         if [ "`echo $i | awk '{ print $1 }'`" = "C" ]; then
>>                 HAS_C=1
>>         fi
>>         echo "A: $HAS_A B: $HAS_B C: $HAS_C"
>> done
>> echo "Final A: $HAS_A B: $HAS_B C: $HAS_C"
>> --------------------------------
>>
>> Result is:
>>
>> --------------------------------
>> $ ./a.sh
>> A: 1 B: 0 C: 0
>> A: 1 B: 1 C: 0
>> A: 1 B: 1 C: 1
>> Final A: 0 B: 0 C: 0
>> --------------------------------
>>
>> It seems to be a bug to me, the envvars lose their values
>> they gained in the loop. It's an ancient bug I must add,
>> I just rechecked it and bash in RedHat 7.1 behaves the same.
>> How can I preserve the variables' values? Putting "export"
>> in front of every assignments doesn't help.
>>
>> Best regards,
>> Zoltán Böszörményi
> 
> 
> I just tried to narrow the test for the bug:
> 
> ---b.sh-------------------------
> #!/bin/bash
> 
> X=0
> A=2
> while [ $A -gt 0 ]; do
>         X=1
>         A=$(($A - 1))
>         echo "A: $A X: $X"
> done
> echo "Final A: $A X: $X"
> --------------------------------
> 
> Result:
> 
> --------------------------------
> $ ./b.sh
> A: 1 X: 1
> A: 0 X: 1
> Final A: 0 X: 1
> --------------------------------
> 
> The variables here keep their values. Hm. Then the bug may come
> from the "cat a.txt | while read i; do" construct, maybe another
> sub-shell is opened for this loop. But then putting "export"
> in front of the assignments inside the loop should solve this but
> it doesn't. It still seems to be a bug to me somewhere.

Yes, the "cat a.txt | while read i; do ..." construct does cause
the "while" loop to be executed in a subshell.  Everything is working
as it should.  The only bug is in your understanding of how processes
and environments work.  Each process has its very own environment, and 
exporting a variable causes a child process's initial enviromnent to
receive a copy of that variable.  Nothing the child subsequently does
to that copy has any effect on the parent process.

Hopefully you've got a copy of the bash FAQ installed on your system
(/usr/share/doc/bash-3.0/FAQ on mine).  Take a look at answer E4.

-- 
Bob Nichols         Yes, "NOSPAM" is really part of my email address.