Front page | perl.perl5.porters |
Postings from October 2015
Karl Williamson wrote: > One solution I thought of (that Zefram) doesn't >like is for F:G to fork a shell if and only if it finds a shell >metacharacter. That would mean that it would acquire the newline misbehaviour, and equivalent misbehaviour on semicolon and other metacharacters. That is not a desirable feature of the glob-via-shell mechanism, it's a problem to be avoided. We should prefer to bring the two globbing mechanisms into synchrony by tweaking the glob-via-shell code to escape non-globbing metacharacters. A problem with this is that the globbing mechanism might be intentionally used to get types of shell expansion other than globbing, such as parameter expansion and command substitution, which don't interfere with the glob-via-shell framework. In the general case these are shell-specific, and can contain arbitrary embedded shell commands, so it's impossible to distinguish metacharacters that are part of these acceptable constructions from metacharacters that will break the framework. But these uses have already been broken by the File::Glob implementation, so it should be fine to break them in the glob-via-shell implementation too. -zeframThread Previous | Thread Next