Re[3]: [cedet-semantic] [Ecb-list] Performance issues

[Eric M. L. <er...@si...>]

>>> "Berndl, Klaus" <kla...@sd...> seems to think that:
>>>Anyway, I didn't even realize that that was in there. I've read the 
>>>documentation, but I must have missed that step. Maybe it should happen 
>>>automatically the first time you start ECB. Is there any reason to NOT byte 
>>>compile the package?
>  [ ... ]
>
>>Here are several reasons I do not byte compile things:
>
>>1) Lazyness
>
>This reason is that one i understand best (no ironic!) :-)
>But for this see answer to 5...
>
>>2) So it works in Emacs and XEmacs from one install area.
>
>If i have understood right the original poster thinks if there are
>any reasons to byte-compile packages AFTER installation...but maybe i'm
>wrong...
>
>>3) Work around byte compiler bugs
>
>Maybe...do not know...

I have some image manipulation macros that cannot be byte compiled
under XEmacs, as the bodies of those functions disappear.

>>4) So you can debug with symbols
>
>Do not understand this...could you please explain this a little
>bit more...AFAIK i can use edebug also very well with byte-compiled
>elisp-code.

Sometimes semantic will do this:

Invalid Function: 1

but I have no idea where it is coming from.  I then run with
`debug-on-error' set to t.  In this case, byte compilation makes the
insides of functions illegible so I don't know how it got there.
With the code loaded, I can often see and fix the given problem w/out
starting edebug.

>>5) So you can modify it regularly and not worry about forgetting to
>>   byte compile it again.
>
>
>The following code prevents you to forget this because it autom.
>byte-compiles if necessary.
>
>(BTW: IMHO such options like 
>mode-specific-after-save-buffer-hooks or any better mechanism
>should be included in Emacs so a user can fast and easy specify
>that changing his elisp-files causes autom. re-byte-compiling
>at save time.)

You should suggest it to the Emacs and XEmacs maintainers.

They would probably respond that a making `after-save-hook'
buffer-local and adding to it in the major-mode hook would be a
better approach.  That would cut down on looping during a save.

>(defvar mode-specific-after-save-buffer-hooks nil
>  "Alist (MAJOR-MODE . HOOK) of after-save-buffer hooks
>specific to major modes.")
>
>(defun run-mode-specific-after-save-buffer-hooks ()
>  "Run hooks in `mode-specific-after-save-buffer-hooks' that match the
>current buffer's major mode.  To be put in `after-save-buffer-hooks'."
>  (let ((hooks mode-specific-after-save-buffer-hooks))
>    (while hooks
>      (let ((hook (car hooks)))
>        (if (eq (car hook) major-mode)
>            (funcall (cdr hook))))
>      (setq hooks (cdr hooks)))))
>
>(add-hook 'after-save-hook 'run-mode-specific-after-save-buffer-hooks)
>
>(defvar byte-compile-not-compile-regexp
>  "\\(\\.emacs\\)\\|\\(\\.gnus\\)\\|\\(subdirs\\.el\\)\\|\\(\\.custom\\.el\\)")
>
>(defun byte-compile-buffer ()
>  "byte-compiles the current-buffer if it is an elisp-buffer"
>  (when (and (eq major-mode 'emacs-lisp-mode)
>             (not (string-match  byte-compile-not-compile-regexp (buffer-name))))
>    (require 'advice)
>    (ad-deactivate-all)
>    (unwind-protect
>        (emacs-lisp-byte-compile)
>      (ad-activate-all))))
>
>(setq mode-specific-after-save-buffer-hooks
>      '((emacs-lisp-mode . byte-compile-buffer)))
  [ ... ]

Nifty idea.  I'd be worried about this being pretty slow.  Perhaps an
idle timer and quit-hook could help.

Eric

-- 
          Eric Ludlam:                 za...@gn..., er...@si...
   Home: http://www.ludlam.net            Siege: www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: www.gnu.org