Re: [boost] Re: relative/absolute paths in filesystem library

From: Beman Dawes (bdawes_at_[hidden])
Date: 2002-11-25 16:35:12

At 03:08 PM 11/25/2002, Justus Schwartz wrote:

 >* Beman Dawes <bdawes_at_[hidden]> [021125 03:26]:
 >> The branch fs_review gives a snapshot of work-in-progress. It isn't
 >stable
 >> at the moment, but progress is definitely being made. I'm hoping to get 
a
 >> lot of work done on it this coming week.
 >
 >i looked into it, as i missed the review. i also looked at the mailing 
list
 >archives but could not find a posting mentioning something about 
symlinks.
 >are
 >there plans to support/handle them? it might be a good idea to change all 

 >the
 >calls to stat in the posix implementation to calls to lstat. and provide
 >functions like is_symlink and symlink_to (i implemented them for a small
 >script i wrote to test the lib)
 >
 >or maybe one could led the user of the lib choose wich behaviour she 
wants
 >(defaulting to handling symlinks as files)
 >
 >sorry if i only missed the messages in the archives and this has all been
 >discussed.

Justus,

Links weren't really discussed before or during the formal review.  Thomas 
Witt and I discussed some aspects of them after the review, and mostly 
deferred hard decisions into the future. We quickly convinced each other 
that behavior in the presence of links needs to be explicitly addressed, 
but neither of us knew the answers to various questions.

One thing we both did agree on was that the presence of links should never 
be allowed to cause pathological behavior, like a program loop. POSIX 
disallows loops (ELOOP error), but I haven't checked other OS's. I'll have 
to do some more research before making any decisions.

Since some operating systems don't even support the concept of links, any 
support in the filesystem library will have to be carefully crafted to work 
portably. Yet at the same time, meet user expectations.

So your questions come at a good time; they need to be resolved in the next 
few weeks.

Thanks,

--Beman