From: Steven J A. <ta...@st...> - 2004-04-27 18:36:21
|
On Tuesday, April 27, 2004, at 08:25 AM, Vince Darley wrote: > On a system (Linux, say) which has a functional realpath(), we would not > want to use your code, so we need a patch which deals with that. man states realpath() differs slightly from the Solaris. A problem in dealing with two instances of realpath() in TclpObjNormalizePath is for example: "/tmp/tcl.foo.dir" The first realpath() is sent "/tmp" which as I have mostly seen should remain as "/tmp". Instead it resolves to "/private/tmp" example:"/tmp/bar" with startAt of 4 This bypasses the first, sends "/tmp" into the second evaluates as "/private/tmp" when it appears the request is to evaluate pass "/tmp" (a relative?). example:"/tmp" This bypasses both realpaths. So should realpath resolve end path entries? should startAt be ignored? man only states that all but the last path must exist. What happened was I used a wrapper around findpath as realpath. The findpath takes a path and resolves based on the understanding that last entry was to be left alone. By splitting into prior + lastdir, these stood out. A little confused. Steve |