Menu
▾
▴
Re: [Matlab-emacs-discuss] Debugging, tab completion in matlab-shell?
|
From: Leo A. <dn...@gm...> - 2010-03-14 00:36:34
|
I am continuing to troubleshoot the issue of errors not turning into
anchors by default in matlab-shell. I got the errors to turn into
matlab by modifying the following variables:
(setq matlab-shell-error-stack-start "^\\?\\?\\?\\s-")
(setq matlab-shell-error-stack-end "^\\s-*$")
In other words, I simply deleted the closing/opening brace after "^"
in the regexp (original definitions were:)
(defvar matlab-shell-error-stack-start "^{\\?\\?\\?\\s-")
(defvar matlab-shell-error-stack-end "^}\\s-*$")
Is this a reasonable fix?.. (I suspect it is not, because now
(string-match matlab-shell-error-stack-end str) on line 4542 seems to
return t for any output, not just errors). However, I am not sure why
the braces were there in the original regular expressions -- it looks
like the only difference between "error" output and normal output is
an extra blank line that appears between the text and the Matlab
command prompt. So perhaps the matlab-shell-error-stack-end regexp
should be set up to match a blank line?..
--Leo
On Tue, Mar 9, 2010 at 12:22 PM, Leo Alekseyev <dn...@gm...> wrote:
> Eric, good catch on comint-prompt-read-only! It was indeed non-nil
> (thanks to some code in EmacsWiki related to ESS). That took care of
> the tab problem (and also fixed a problem when interacting with R
> buffers through ESS).
>
> However, my errors still don't turn into links -- even running with
> emacs -q -- and the behavior where the last actual hyperlink (e.g.
> from running help) is assumed to be the last error (when doing M-g n)
> still persists. Any ideas?.
>
> --Leo
>
> On Tue, Mar 9, 2010 at 11:01 AM, Eric Ludlam <Eri...@ma...> wrote:
>> Do you perhaps have the variable `comint-prompt-read-only' set to some non-nil value?
>>
>> Older Emacs versions used to have this be non-nil, but I was pretty sure I had fixed matlab.el for this, but things may have regressed while I was using newer versions of Emacs.
>>
>> If this is not the case, you can try starting Emacs with "-q", to not load your init file, and then hand-load in matlab.el, and try your test case. If that works, then there is some configuration issue in your .emacs file.
>>
>> Eric
>>
>> -----Original Message-----
>> From: leo...@gm... [mailto:leo...@gm...] On Behalf Of Leo Alekseyev
>> Sent: Monday, March 08, 2010 7:17 PM
>> To: Eric Ludlam
>> Cc: mat...@li...
>> Subject: Re: [Matlab-emacs-discuss] Debugging, tab completion in matlab-shell?
>>
>> Eric,
>> Thanks for your response.
>>
>> Regarding matlab-shell-collect-command-output -- it prints the
>> following to the messages buffer: save-excursion: Text is read-only.
>> So it looks like for some reason it fails to delete the output and
>> bails.
>>
>> I have no idea what's going on with errors not being converted to
>> links -- but perhaps something similar is preventing the code from
>> modifying the buffer?..
>>
>> But the matlab shell buffer itself is not read-only (buffer-read-only
>> is nil; using (barf-if-buffer-read-only) has no effect...)
>>
>> Any ideas on what might be happening?
>>
>> --Leo
>>
>> On Mon, Mar 8, 2010 at 8:43 AM, Eric Ludlam <Eri...@ma...> wrote:
>>> Hi,
>>>
>>> I gave this a quick run and found that it was working as I expected still, and I have the CVS version of matlab.el, with no pending changes.
>>>
>>> When "tab" is pressed, it should split the window, and display the completions in a separate buffer, independent of the matlab-shell buffer. It will also do the completion if there is only one option. The misc text you quoted below is what matlab-shell does under-the-covers to get the completion list.
>>>
>>> It works internally by flipping into a "capture some text" mode, then back out again when the capture is done, or when another prompt ">>" is shown again. K prompts "K>>" and student prompts "EDU>>" should also work.
>>>
>>> It will also key off of carriage returns, and a few other things to try and distinguish between text to be captured, text to turn into links, and regular text.
>>>
>>> I'll take a guess that some bit of text has put it into a bad state. Here are some possibilities:
>>>
>>> If you have some non-standard prompt, see line 4392 where comint-prompt-regexp is set. There are a few other hard-coded prompts around too.
>>>
>>> If the error text is somehow "new", or contains bad text, look at matlab-shell-render-errors-as-anchor, or matlab-shell-render-html-as-anchor, and the regular expressions that go into those.
>>>
>>> Lastly, there is matlab-shell-collect-command-output, which captures output, which sends a command, then spins waiting for a prompt to come up. This is exiting early w/ your tab completion. Perhaps there is an issue with carriage returns?
>>>
>>> Good Luck
>>> Eric
>>>
>>> -----Original Message-----
>>> From: Leo Alekseyev [mailto:dn...@gm...]
>>> Sent: Friday, March 05, 2010 11:22 PM
>>> To: mat...@li...
>>> Subject: [Matlab-emacs-discuss] Debugging, tab completion in matlab-shell?
>>>
>>> Hi all,
>>> I recently started using matlab-emacs (with matlab-shell) on a more
>>> regular basis, and I am reasonably happy with the workflow -- but I am
>>> also seeing some anomalies, in particular with debugging and
>>> tab-completion.
>>>
>>> First, if I encounter an error during execution, the offending file
>>> name does not turn into a link (like it does using Matlab GUI), which
>>> I believe should also be happening here. This also appears to break
>>> the "go to last error" functionality: if there are links anywhere else
>>> in the command window (e.g. as a result of "help" command) the "go to
>>> last error" will follow the link e.g. to the documentation page,
>>> ignoring the error. (If there aren't links, then it will jump to the
>>> error line in the offending file, but if I've run "help" on any
>>> command, the go-to-last-error is now broken for this session). I
>>> would really appreciate pointers on how I might troubleshoot this --
>>> the behavior is clearly broken.
>>>
>>> Second (minor) issue is that tab completion (e.g. typing test<tab>)
>>> dumps the following to the buffer:
>>>>> matlabMCRprocess = com.mathworks.jmi.MatlabMCR
>>> matlabMCRprocess =
>>> com.mathworks.jmi.MatlabMCR@7deb41d6
>>> >> matlabMCRprocess.mtFindAllTabCompletions('test'), clear('matlabMCRprocess');
>>> ans =
>>> java.lang.String[]:
>>> 'test'
>>> 'test_step_by_step'
>>> 'test_trainlm_gradient'
>>> 'testm'
>>> 'testobjectiveinfo'
>>> 'testscript
>>>
>>> It this the intended behavior?.. Would it be possible to just display
>>> stuff after java.lang.String[]?..
>>>
>>> Relevant versions: GNU Emacs 23.1.50.1 (x86_64-pc-linux-gnu), Matlab
>>> 7.8.0.347 (R2009a) 64-bit (glnxa64), latest version of matlab-emacs
>>> from CVS.
>>> I'd appreciate any pointers regarding the above!
>>> Thanks,
>>> --Leo
>>>
>>> ------------------------------------------------------------------------------
>>> Download Intel® Parallel Studio Eval
>>> Try the new software tools for yourself. Speed compiling, find bugs
>>> proactively, and fine-tune applications for parallel performance.
>>> See why Intel Parallel Studio got high marks during beta.
>>> http://p.sf.net/sfu/intel-sw-dev
>>> _______________________________________________
>>> Matlab-emacs-discuss mailing list
>>> Mat...@li...
>>> https://lists.sourceforge.net/lists/listinfo/matlab-emacs-discuss
>>>
>>
>
|
