bug#15231: 24.3.50; eshell: variable `name' treated different from other variables

bug#15231: 24.3.50; eshell: variable `name' treated different from other variables

[Gregor Zattler]

Dear Emacs developers,

eshell treats a variable named `name' different than a variable
named `foo' (or any(?) other name):

How to reproduce:

emacs -Q -nw -f eshell
#now type
for foo in * { echo "$foo" } 
# this lists the directory Emacs was started in
#now type
for name in * { echo "$name" } 
# this instead lists as many times 'name' as the above
# for statement listed objects in the directory.

E.g.:

~/src/emacs/doc $ for foo in * { echo "$foo" } 
Emacs/
lispintro/
lispref/
man/
misc/
~/src/emacs/doc $ for name in * { echo "$name" } 
name
name
name
name
name
~/src/emacs/doc $


This is the case  with 

GNU Emacs 24.3.50.1 (i486-pc-linux-gnu, X toolkit, Xaw scroll
bars) of 2013-08-31 on dex, modified by Debian

and with

GNU Emacs 23.4.1 (i486-pc-linux-gnu, X toolkit, Xaw3d scroll
bars) of 2013-07-01 on binet, modified by Debian


Is this a bug?

Thanks, Gregor
 

bug#15231: 24.3.50; eshell: variable `name' treated different from other variables

[Stefan Monnier]

> Is this a bug?

Sounds like it, yes.  More specifically it sounds like changing Eshell
to use lexical-binding would fix this bug.


        Stefan "just reading his crystal ball"

Re: bug#15231: 24.3.50; eshell: variable `name' treated different from other variables

[Aidan Gauland]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Is this a bug?
>
> Sounds like it, yes.  More specifically it sounds like changing Eshell
> to use lexical-binding would fix this bug.

But why is this happening with just `name'?

bug#15231: 24.3.50; eshell: variable `name' treated different from other variables

[Stefan Monnier]

>>> Is this a bug?
>> Sounds like it, yes.  More specifically it sounds like changing Eshell
>> to use lexical-binding would fix this bug.
> But why is this happening with just `name'?

I don't know the precise cause of the bug, but my crystal ball says that
the Eshell code that interprets your command itself uses a variable
named "name".  There are probably a few other variables names used by
Eshell that suffer from the same kind of "capture".


        Stefan

bug#15231: 24.3.50; eshell: variable `name' treated different from other variables

[Stefan Monnier]

> I don't know the precise cause of the bug, but my crystal ball says that
> the Eshell code that interprets your command itself uses a variable
> named "name".  There are probably a few other variables names used by
> Eshell that suffer from the same kind of "capture".

I've just tried to convert Eshell to use lexical-binding and it appears
that this problem would indeed be fixed by such a change (if someone has
some kind of test suite for Eshell it would be very helpful, since
changing it to use lexical-binding is not completely trivial and might
introduce unexpected regressions).


        Stefan

bug#15231: 24.3.50; eshell: variable `name' treated different from other variables

[Michael Albinus]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> I've just tried to convert Eshell to use lexical-binding and it appears
> that this problem would indeed be fixed by such a change (if someone has
> some kind of test suite for Eshell it would be very helpful, since
> changing it to use lexical-binding is not completely trivial and might
> introduce unexpected regressions).

There is test/eshell.el. Btw, shouldn't it be moved to test/automated/ ?

>         Stefan

Best regards, Michael.

bug#15231: 24.3.50; eshell: variable `name' treated different from other variables

[Stefan Monnier]

>> I've just tried to convert Eshell to use lexical-binding and it appears
>> that this problem would indeed be fixed by such a change (if someone has
>> some kind of test suite for Eshell it would be very helpful, since
>> changing it to use lexical-binding is not completely trivial and might
>> introduce unexpected regressions).
> There is test/eshell.el. Btw, shouldn't it be moved to test/automated/ ?

Cool, indeed.  I moved the file, added a test to catch this bug, and
changed Eshell to use lexical-binding.


        Stefan