On 4/25/06, Axel Liljencrantz <[EMAIL PROTECTED]> wrote:
> On 4/25/06, Nick Pilon <[EMAIL PROTECTED]> wrote:
> > On 4/25/06, Axel Liljencrantz <[EMAIL PROTECTED]> wrote:
> > > On 4/25/06, Nick Pilon <[EMAIL PROTECTED]> wrote:
> > > > If you try to add a directory that doesn't exist to your PATH using a
> > > > copy of fish built from the current darcs source, you get a whole lot
> > > > of confusing errors:
> > > >
> > > > set: Could not add component /foo to PATH.
> > > > set: Value too large to be stored in data type
> > > >
> > > > Not only does this not explain what the problem is, but you get it
> > > > repeated a whole bunch of times. (Every time fish launches a subshell,
> > > > I think) The Right Thing To Do would seem to be to have a sensible
> > > > error message (set: Directory does not exist) displayed exactly once.
> > >
> > >
> > > On my system:
> > >
> > > fish> set PATH $PATH /fdsa
> > > set: Could not add component /fdsa to PATH.
> > > set: No such file or directory
> > >
> > > E.g it works exactly as you describe. Any insights into why this is
> > > broken on OS X is welcome.
> >
> > Found the problem. Somewhere between builtin_set.c:507 and
> > builtin.c:154, errno is getting changed from "2" (no such file or
> > directory) to "84" (value too large). There are a /lot/ of function
> > calls between these two lines, and I believe that relying on errno to
> > remain unchanged across all of them is very bad practice in general.
> > What the code should be doing is saving the errno after it detects the
> > error, then using its saved version (perhaps as a parameter to
> > builtin_wperror ?) instead of the global errno.
> >
> > Want me to patch it to do this? There only appear to be about five
> > calls to builtin_wperror, so it doesn't look too time-consuming.
>
> Thanks for checking this out. A patch would be most welcome, but I'm
> wondering if that is the only problem. Didn't you say the error
> message gets printed repeatedly? If so, that implies that something
> more is going wrong.
Oh, yes. If you have an invalid entry in your PATH when fish loads, it
gets printed a whole bunch of times. The first time you do a
tab-completion with an invalid entry in your PATH, you get a bunch
more - well more than a screenful. I'm not sure what's doing that -
perhaps subshell invocation?
--
-Nick Pilon
Rȧ�:&q�[���y�hv����^y�h��i��py� ���z� r���!���n}�h�ꮉ�%����ފ{^���y�^r薈2����쨺��m欉�ã
塧HŞm*az����bq�b�t�����]5m� v����!xg��x��m���zV���ږF�����\�
