Should C-c C-k try to load a resulting fasls file ALWAYS?
Should it only do so if COMPILE-FILE did not return T as failurep?
From glancing, SBCL's and CMUCL's swank backends do the latter.
Most other backends use an implementation dependent :LOAD keyword argument to COMPILE-FILE which I haven't checked what their semantics are on this matter. I wouldn't be surprised if the differ.
The thing is that I sometimes want the first behaviour (early stage of development), sometimes the latter (precious state in the image).
If C-c C-k always loaded, you could get the other behaviour by using C-c M-k instead (compile only) and C-c C-l afterwards if compilation didn't fail.
If C-c C-k regarded failurep, you could force loading afterwards by C-c C-l.
So both behaviours are available. I just pledge for consistency across backends.
-T.
On 21 August 2010 11:03, Tobias C Rittweiler tcr@freebits.de wrote:
Should C-c C-k try to load a resulting fasls file ALWAYS?
Should it only do so if COMPILE-FILE did not return T as failurep?
From glancing, SBCL's and CMUCL's swank backends do the latter.
...which I've been bitten a couple of times by. :)
How about: pop up the debugger with a restart to load the file if failurep is true?
Cheers,
-- Nikodemus
* Nikodemus Siivola [2010-09-02 11:59] writes:
On 21 August 2010 11:03, Tobias C Rittweiler tcr@freebits.de wrote:
Should C-c C-k try to load a resulting fasls file ALWAYS?
Should it only do so if COMPILE-FILE did not return T as failurep?
From glancing, SBCL's and CMUCL's swank backends do the latter.
...which I've been bitten a couple of times by. :)
How about: pop up the debugger with a restart to load the file if failurep is true?
The problem is that compiler warnings often contain the most interesting information and popping up a debugger either explicitly or by load-time code hides the compiler messages (actually, aborting the debugger also aborts the part that sends the warnings to Emacs so the user doesn't see the warnings).
I think it would be better to display the compiler warnings first and then ask the user (with some y-or-n style query) whether the fasl file should be loaded. Essentially move the decision to the Emacs side; that way customization would also be easier.
Helmut
On 2 September 2010 18:26, Helmut Eller heller@common-lisp.net wrote:
I think it would be better to display the compiler warnings first and then ask the user (with some y-or-n style query) whether the fasl file should be loaded. Essentially move the decision to the Emacs side; that way customization would also be easier.
This suits me perfectly -- as long as the user (1) knows code wasn't loaded yet (2) is able to force a load, I'm more than happy.
Cheers,
-- Nikodemus
-
Helmut Eller -
Nikodemus Siivola -
Tobias C Rittweiler