This 'fix' doesn't work on some systems. I'm not sure which system caused problems for Costin, but it fails on a Mac.
File.getCanonicalPath() is system dependant. In particular, on a Mac, the filename returned ends in a '/'. This is different to the behaviour of File.getAbsolutePath(). Hence, these will never be equal on a mac. You can fix this by normalizing for the existance of a '/' at the end of the paths.
Also, the simple existance test shown below tests if there is a symlink anywhere in the path, NOT if the path target is a symlink. I believe there was other code elsewhere that got the canonical path when you started to remove all other symlinks but, based on Costin's report, this didn't work all the time.
later,
\x/ill :-}
--On Tuesday, April 4, 2000 12:11 AM -0700 [EMAIL PROTECTED] wrote:
jon * writes: > on 4/2/00 10:27 PM, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote: > > > I can't convince cvs to allow me to log in (anonymously) to get the > > latest source, but at one time this also happened to me. I submitted > > a patch, which tested to see if the "directory" was a symlink, and > > ignored it if it was. The patch was committed by someone, and all > > seemed well. > > submit the patch again...i will get it checked in. > > really this conversation should be going on the ant-dev mailing list.
I looked at the source via cvsweb (I wonder why my logins time out with "regular" anonymous cvs?) and noticed that costin commented out the "fix"
// if (dir.getCanonicalPath().equals(dir.getAbsolutePath())) { // (costin) It will not work if /home/costin is symlink to /da0/home/costin ( taz // for example )
So, it looks as though the dangerous behaviour of Deltree has returned.
-- Jay Doane | [EMAIL PROTECTED]
