Re: [Matlab-emacs-discuss] Missing execution line pointer after upgrading Matlab

[Eric L. <Eri...@ma...>]

Hi,

That looks like a great idea for a simple fix.   You might be able to shadow “dbstep” in the Matlab-emacs toolbox directory, and have it call

builtin(‘dbstep’,...)

to dispatch, then fprintf items out.

I’m not sure why it wouldn’t remove the hotlink text.   That will need some time in the Emacs debugger around matlab-shell-render-anchor to know.

Eric

From: Odd Andersen [mailto:odd...@gm...]
Sent: Friday, January 29, 2016 12:10 PM
To: mat...@li...
Subject: Re: [Matlab-emacs-discuss] Missing execution line pointer after upgrading Matlab

Hi Eric, and thanks for answering.
I have looked a bit at the problem myself.
I was able to find a relatively nonintrusive (but hacky) workaround by writing a function that emulates the "bug"/"feature" of earlier Matlab versions, putting it in the "toolbox" folder, and calling it alongside with the db-commands that are specified for gud in matlab.el.
In other words, placing the following file in 'toolbox':
function dbhotlink()
   [ST, I] = dbstack('-completenames');
   fprintf('<a href="matlab: opentoline(''%s'',%i,1)">%i </a>\n', ST(2).file, ST(2).line, ST(2).line);
end

and then modifying lines 4542 and 4543 in matlab.el to read:
(gud-def gud-step   "dbstep in;\ndbhotlink();\n"           "\C-s" "Step one source line, possibly into a function.")
(gud-def gud-next   "dbstep %p;\ndbhotlink();\n"           "\C-n" "Step over one source line.")
This sort of works, but with some drawbacks:
- The text line generated by 'dbhotlink' is not properly removed from the pty (could perhaps be fixed with some modification of the relevant regexes..)
- The stack pointer is only updated when using the "next" (C-n) or "step in" (C-s) functions.
Odd A.
[Image removed by sender.]

On Fri, Jan 29, 2016 at 6:08 PM, Odd Andersen <odd...@gm...<mailto:odd...@gm...>> wrote:
Hi Eric, and thanks for answering.
I have looked a bit at the problem myself.
I was able to find a relatively nonintrusive (but hacky) workaround by writing a function that emulates the "bug"/"feature" of earlier Matlab versions, putting it in the "toolbox" folder, and calling it alongside with the db-commands that are specified for gud in matlab.el.
In other words, placing the following file in 'toolbox':
function dbhotlink()
   [ST, I] = dbstack('-completenames');
   fprintf('<a href="matlab: opentoline(''%s'',%i,1)">%i </a>\n', ST(2).file, ST(2).line, ST(2).line);
end

and then modifying lines 4542 and 4543 in matlab.el to read:
(gud-def gud-step   "dbstep in;\ndbhotlink();\n"           "\C-s" "Step one source line, possibly into a function.")
(gud-def gud-next   "dbstep %p;\ndbhotlink();\n"           "\C-n" "Step over one source line.")
This sort of works, but with some drawbacks:
- The text line generated by 'dbhotlink' is not properly removed from the pty (could perhaps be fixed with some modification of the relevant regexes..)
- The stack pointer is only updated when using the "next" (C-n) or "step in" (C-s) functions.
Odd A.

On Fri, Jan 29, 2016 at 3:30 PM, Eric Ludlam <Eri...@ma...<mailto:Eri...@ma...>> wrote:
Hi,

I have looked into this, and in Earlier MATLABs I used to take advantage of a “bug”, which I thought was a feature, that allowed MATLAB to send me html like anchors which I could interpret.  MATLAB is now correctly identifying that it is not pointing at a device that supports anchors (since it is a plain pty), so I don’t have the data.

I asked a couple folks about this and I don’t believe there is a trick I can pull to get this data to be output to Emacs with the newer MATLABs, which leaves things kind of broken.

Before the links, I used to interpret the plaintext, but this was buggy, and simple output tweaks from MATLAB, or locale changes would mess it up, so it doesn’t seem too good either, but could be resurrected.  I don’t really have the bandwidth for that kind of refactoring project though. ☹

Eric

From: Odd Andersen [mailto:odd...@gm...<mailto:odd...@gm...>]
Sent: Wednesday, January 27, 2016 8:21 AM
To: mat...@li...<mailto:mat...@li...>
Subject: [Matlab-emacs-discuss] Missing execution line pointer after upgrading Matlab


Hi,
After I upgraded my Matlab version from R2014a to R2015b, I can no longer see the little arrow that indicates the current execution line while debugging matlab code using matlab-emacs.  Moreover, debugging seems to work poorly in other ways: it apparently no longer brings up the relevant code buffer while debugging (possibly tied to the problem above), and when arriving at a multi-line function call (i.e. where each line of text is continued with '...'), it no longer treats this as one line but insists on stepping over each text line separately.
I saw someone else (Tunc Aydin) mentioned the missing execution line pointer on this mailing list last fall, and was answered by Eric Ludlam that apparently the "output Emacs was interpreting for the debugger has changed" and that this .  (http://sourceforge.net/p/matlab-emacs/mailman/message/34601964/)
Has anyone looked into this since then?  matlab-emacs has worked wonders for me in the past, and it's very sad if it stops providing a practical debugging tool  as people upgrade to newer versions of Matlab.
Best,
Odd Andersen