more useful options | |
PerlMonks |
comment on |
( [id://3333]=superdoc: print w/replies, xml ) | Need Help?? |
I've just spent more than one hour debugging very wierd problem. Test script for one of my projects I'm currently working on died with error like:
Let's look in Net/DNS/Resolver.pm: As you can see on load of Net::DNS::Resolver module it tries to read DNS resolver configuration file in subroutine read_config() if it runs on Unix and there exist this file. Nothing wrong with it at first look. Yet it fails with 'Modification of a read-only value attempted' error on while (<FILE>) { line. Turned out that I had some code which could load Net::DNS module on demand and this code was called inside of map. My next step was finding that following snipplet dies with same error. map { require Net::DNS } 1; Than I removed all non-essential details and I got final test case: map { while(<>) { } } 1; Suddendly this error started to make sense to me. map associates $_ to elements of the passed list. In this case it is a list with a single element - constant 1. As it is a constant $_ becomes read-only but while(<>) { } still tries to assign data obtained from file handle to $_ and it results in this error. So what is the conclusion? Do not use while(<>) { } unless you localize $_ or you are completly sure that nobody will call you code from map (or grep, or for/foreach - they are subject to the same problem). Personally I prefer always using while(my $line = <>) construct. Update: These nodes demonstrate same problem with while(<>) { ... }: Method call to tied hash leads to file read error and opening a file destroys nulling entries in a list?!?!.
-- In reply to while(<>) { ... } considered harmful by IlyaM
|
|