precedence of auto-mode-alist and magic-fallback-mode-alist

precedence of auto-mode-alist and magic-fallback-mode-alist

[Stephen Berman]

On GNU Emacs 23.0.50.3 (i686-pc-linux-gnu, GTK+ Version 2.12.0) of
2007-11-29 on escher I do the following:

1. emacs -Q
2. M-x browse-url-emacs RET http://www.gnu.org RET
3. After a short time a buffer www.gnu.org pops up containing the HTML
source of the gnu.org home page.  This buffer is in Org mode, because
auto-mode-alist takes precedence over magic-fallback-mode-alist (if I
save the buffer as "testme" then C-x C-f opens the file in html-mode).

I don't remember if cases like this were considered when this precedence
relation was established.  If not, is the above an argument against this
precedence relation, or just evidence that we can't always DTRT, or is
it a bug that can be fixed while retaining the current precedence
relation?

Steve Berman

Re: precedence of auto-mode-alist and magic-fallback-mode-alist

[Lennart Borgman (gmail)]

Stephen Berman wrote:
> On GNU Emacs 23.0.50.3 (i686-pc-linux-gnu, GTK+ Version 2.12.0) of
> 2007-11-29 on escher I do the following:
> 
> 1. emacs -Q
> 2. M-x browse-url-emacs RET http://www.gnu.org RET
> 3. After a short time a buffer www.gnu.org pops up containing the HTML
> source of the gnu.org home page.  This buffer is in Org mode, because
> auto-mode-alist takes precedence over magic-fallback-mode-alist (if I
> save the buffer as "testme" then C-x C-f opens the file in html-mode).
> 
> I don't remember if cases like this were considered when this precedence
> relation was established.  If not, is the above an argument against this
> precedence relation, or just evidence that we can't always DTRT, or is
> it a bug that can be fixed while retaining the current precedence
> relation?
> 
> Steve Berman


Should not the rules be different for a file fetched from the net? We 
can't know association between the URL and the content type. (Maybe the 
headers can be used?)

BTW, why is the new-window argument ignored in browse-url-emacs?

Re: precedence of auto-mode-alist and magic-fallback-mode-alist

[Stefan Monnier]

>> 2. M-x browse-url-emacs RET http://www.gnu.org RET
>> 3. After a short time a buffer www.gnu.org pops up containing the HTML
>> source of the gnu.org home page.  This buffer is in Org mode, because
>> auto-mode-alist takes precedence over magic-fallback-mode-alist (if I
>> save the buffer as "testme" then C-x C-f opens the file in html-mode).
>> 
>> I don't remember if cases like this were considered when this precedence
>> relation was established.  If not, is the above an argument against this
>> precedence relation, or just evidence that we can't always DTRT, or is
>> it a bug that can be fixed while retaining the current precedence
>> relation?

> Should not the rules be different for a file fetched from the net?

Indeed.  We should use the content-type info provided in the HTTP
response, but we don't have any support for that right now.


        Stefan

Re: precedence of auto-mode-alist and magic-fallback-mode-alist

[Jason Rumney]

Stephen Berman wrote:
> On GNU Emacs 23.0.50.3 (i686-pc-linux-gnu, GTK+ Version 2.12.0) of
> 2007-11-29 on escher I do the following:
>
> 1. emacs -Q
> 2. M-x browse-url-emacs RET http://www.gnu.org RET
> 3. After a short time a buffer www.gnu.org pops up containing the HTML
> source of the gnu.org home page.  This buffer is in Org mode, because
> auto-mode-alist takes precedence over magic-fallback-mode-alist (if I
> save the buffer as "testme" then C-x C-f opens the file in html-mode).
>
> I don't remember if cases like this were considered when this precedence
> relation was established.  If not, is the above an argument against this
> precedence relation, or just evidence that we can't always DTRT, or is
> it a bug that can be fixed while retaining the current precedence
> relation?
>   

The word fallback in the variable name suggests that it is intended to
work this way. There is a corresponding magic-mode-alist which has
higher priority. But the real bug is that www.gnu.org is being treated
as a filename after being fetched by browse-url. It shouldn't be going
anywhere near auto-mode-alist.