mailing-list for TeXmacs Users

[Nicola Mingotti <address@hidden>]

Hello Max !

So, the most difficult part was for me to discover how to recompile 
efficiently the FreeBSD port.
https://forums.freebsd.org/threads/how-to-tweak-the-source-of-an-existing-port.69164/#post-413749

Trying to get the result i was bit messy, sorry about that.

I left there the ::killpg because i had no idea if it had any side effect,
besides throwing a signal (which is wrong: negative).
I added the standard "kill", instead of the "::kill" because I was more
confident it was doing what i wanted, since i don't control Qt.

Anyhow, today i used your code in the patch and it all seems to work :)
The patch I attach contains your lines more a STDOUT message.

I left the message in STDOUT because it is important to know the signal
is sent. It is the plugins that at times do not detect it AFAIK-say.
(eg. Python plugin does not get SIGINT, R plugin does)

Bye
nicola

==== Patching procedure for the FreeBSD port : editors/texmacs 
======================
==== tested in FreeBSD-11.2

1] It is supposed you installed the port system

2] Remove TeXmacs package, if you have it installed via package
#> pkg remove texmacsXXX

3] Get the source, clean if necessary
$> cd /usr/ports/editors/texmacs
# if necessary
$> sudo make deinstall
$> sudo make clean
$> sudo make fetch

4] Put the patch file (attached to this mail) into
/usr/ports/editors/texmacs/files/patch-src_Plugins_Qt_qt__pipe__link.cpp

5] patch the source and compile
# in config i keep checked all options
$> sudo make config
$> sudo make patch
$> sudo make build

6] check that it is working
$> cd /usr/ports/editors/texmacs/work/TeXmacs-1.99.4-src/TeXmacs
$> export TEXMACS_PATH=`pwd`
$> cd bin
$> ./texmacs
# now install Max test plugin and and restart TeXmacs
# test that that "interrupt" is resetting the counter !
# not only that there is written "Interrupted"
# You should see a message in STDOUT sayin the singnal has beeen sent.

7] install the software
$> cd /usr/ports/editors/texmacs
$> sudo make install

=========================================================










On 1/16/19 12:32 AM, Massimiliano Gubinelli wrote:
> Dear Nicola,
>   I do not understand the patch. You need to send *two* kill (first 
> ::killpg() and then killI()) to the same pid? or is putting the pid in the 
> local variable that makes things works?
>
> Is not clear to me what is going on.
>
> Best
> Max
>
>
>
> > On 16. Jan 2019, at 01:21, Nicola Mingotti <address@hidden> wrote:
> >
> >
> > I was able to fix it !
> >
> > After patching, my Ruby plugin and Max test plugin both work.
> >
> > Without the patch my plugin does not get the SIGINT and Max plugin
> > does not reset the counter.
> >
> > The working patch for TeXmacs 1.99.4 (tested valid for FreeBSD 11.2) is :
> >
> > -----------------------
> > --- src/Plugins/Qt/qt_pipe_link.cpp.orig        2019-01-15 23:10:27 UTC
> > +++ src/Plugins/Qt/qt_pipe_link.cpp
> > @@ -9,6 +9,9 @@
> >   * in the root directory or <http://www.gnu.org/licenses/gpl-3.0.html>.
> >   
> > ******************************************************************************/
> >
> > +// for my debug
> > +#include<iostream>
> > +
> >   #include "basic.hpp"
> >
> >   #if defined (QTTEXMACS) && (defined (__MINGW__) || defined (__MINGW32__) || 
> > defined (QTPIPES))
> > @@ -144,8 +147,15 @@ qt_pipe_link_rep::interrupt () {
> >     if (!alive) return;
> >   #if defined(__MINGW__) || defined(__MINGW32__)
> >     // Not implemented
> > -#else
> > +#else
> >     ::killpg(PipeLink.pid (), SIGINT);
> > +
> > +  // for some reason the prevoius "kill" kills
> > +  // (-1)*myPid.
> > +  int myPid = PipeLink.pid();
> > +  cout << "DBG> signal SIGINT on pid: " << myPid << "\n";
> > +  kill(myPid, SIGINT);
> > +
> >   #endif
> >   }
> > -----------------------
> >
> > bye
> > Nicola
> >
> >
> >
> >
> > On 1/15/19 12:35 PM, Nicola Mingotti wrote:
> > > Please don't put much wait on what i wrote,
> > > I may have made a mistake in recompiling the FreeBSD port
> > > which messed all the thing up.
> > >
> > > The negative PID is there, but my changes to the code may
> > > have not be taken at all in subsequent tests ... i am checking and
> > > trying to understand.
> > >
> > > I will let you know ASAP.
> > >
> > > sorry
> > > bye
> > > n.
> > >
> > > On 1/14/19 5:39 PM, Nicola Mingotti wrote:
> > > > Hello Max,
> > > >
> > > > I have had finally time to do further experiments.
> > > >
> > > > I tried your test plugin, in my system (FreeBSD 11.2, Texmacs v. 1.99.4 from 
> > > > ports) it works
> > > > with and without your patch. Quite misteriously i must say.
> > > >
> > > > My plugin instead keeps on not working and not receiving SIGINT.
> > > >
> > > > I start to doubt i have a bug in my code somewhere, but it is wired,
> > > > if I run my plugin on the console it behaves well on Ctrl-C !
> > > >
> > > > I have discovered something that is interesting.
> > > >
> > > > If I run:
> > > > $> truss ./texmacs -x '(make-session "ruby" "default")' 2>&1 | grep '^kill'
> > > >
> > > > I get this stuff :
> > > > kill(-59846,SIGINT)   # ERR3, no such process
> > > >
> > > > ... i have never seen a negative pid in life.
> > > >
> > > > I tried to change TeXmacs code you modified many times, but it seems the damn 
> > > > negative
> > > > pid not going to change, whatever i do, eg :
> > > > ---
> > > >    Q_PID pid = PipeLink.pid ();
> > > >    int ret =  ::kill(-pid, SIGINT);
> > > > ---
> > > >
> > > > A different but still negative pid appears to be sent to your plugin,
> > > > but, for some reason, your plugin gets the SIGINT.
> > > >
> > > > I am extremely puzzled,
> > > > do you have any idea ?
> > > >
> > > > bye
> > > > n.
> > > >
> > > >
> > > > On 12/12/18 8:56 AM, Massimiliano Gubinelli wrote:
> > > > > Hi again,
> > > > >
> > > > > > On 11. Dec 2018, at 20:11, Nicola Mingotti <address@hidden> wrote:
> > > > > >
> > > > > > Hi again,
> > > > > >
> > > > > > I am puzzled, I am trying to get the SIGINT from my plugin
> > > > > > but it seems to me TeXmacs is not sending that signal.
> > > > > Seems there is a bug with the signals. I attach a patch and a test program 
> > > > > (which has to be installed as a plugin).
> > > > >
> > > > > I will commit a patch in svn.
> > > > >
> > > > > Best
> > > > > Max
> > > > >
> > > > >
> > > > > -----------------------------------------------------------------------
> > > > > The patch
> > > > > -8<---------------------------------------------------------------------
> > > > > Index: src/Plugins/Qt/qt_pipe_link.cpp
> > > > > ===================================================================
> > > > > --- src/Plugins/Qt/qt_pipe_link.cpp    (revision 11316)
> > > > > +++ src/Plugins/Qt/qt_pipe_link.cpp    (working copy)
> > > > > @@ -148,7 +148,11 @@
> > > > >      qt_error << "SIGINT not implemented on Windows\n";
> > > > >    #else
> > > > >      Q_PID pid = PipeLink.pid ();
> > > > > -  int ret =  ::killpg (pid, SIGINT);
> > > > > +
> > > > > +  // REMARK: previously there were here below a call to ::killpg which does 
> > > > > not seems to work on MacOS
> > > > > +  // I (mgubi) replaced it with ::kill which does the job. But I do not 
> > > > > undestand the difference.
> > > > > +
> > > > > +  int ret =  ::kill (pid, SIGINT);
> > > > >      if (ret == -1) {
> > > > >        qt_error << "Interrupt not successful, pid: " << pid << " return code: " << errno 
> > > > > << "\n";
> > > > >      }
> > > > > -8<--------------------------------------------------------------------
> > > > > The test program
> > > > > -8<--------------------------------------------------------------------
> > > > > #include <iostream>
> > > > > #include <unistd.h>
> > > > > #include <csignal>
> > > > >
> > > > > using namespace std;
> > > > >
> > > > > #define DATA_BEGIN   ((char) 2)
> > > > > #define DATA_END     ((char) 5)
> > > > > #define DATA_ESCAPE  ((char) 27)
> > > > >
> > > > > bool trig = true;
> > > > > int cc = 0;
> > > > >
> > > > > void signalHandler( int signum ) {
> > > > >       trig = false;
> > > > > }
> > > > >
> > > > > int
> > > > > main () {
> > > > >     cout << DATA_BEGIN << "verbatim:";
> > > > >     cout << "STARTING\n";
> > > > >     cout << DATA_END;
> > > > >     cout.flush ();
> > > > >
> > > > >     // register signal SIGINT and signal handler
> > > > >     signal(SIGINT, signalHandler);
> > > > >
> > > > >       while (1) {
> > > > >           char buffer[100];
> > > > >           cin.getline (buffer, 100, '\n');
> > > > >           cout << DATA_BEGIN << "verbatim:";
> > > > >           cout << "Input: " << buffer << "\n";
> > > > >           cout.flush ();
> > > > >
> > > > >           while (trig) {
> > > > >               cout << "Going to sleep.... (" << cc ++ << ")\n";
> > > > >               cout.flush ();
> > > > >               sleep (1);
> > > > >           }
> > > > >           cout << DATA_END;
> > > > >           cout.flush ();
> > > > >           trig = true; cc = 0;
> > > > >       }
> > > > >     return 0;
> > > > > }
> > > > > -8<--------------------------------------------------------------------
> > > > >
> > > > >
> > > > > > I am using TeXmacs version 1.99.4 in FreeBSD.
> > > > > >
> > > > > > Do you know of any plugin who is actually managing SIGINT correctly ?
> > > > > > I tried Python (not working, not implemented IFACSay), Scheme (no button), 
> > > > > > Maxima (not working).
> > > > > >
> > > > > > I test the SIGINT is not working in this way, for Python i can actually
> > > > > > check there is not signal receiver in the code, so, not a TeXmacs problem, 
> > > > > > but the method I use in general is:
> > > > > >
> > > > > > 1] Run a sleep command, 10 seconds
> > > > > > 2] Press the STOP button in TeXmacs
> > > > > > 3] Run a simple arithmentic command e.g. 1+1
> > > > > >
> > > > > > === Pyjhon ===
> > > > > > > import time
> > > > > > > timp.sleep(10)
> > > > > > PRESS QUIT
> > > > > > > 1+1
> > > > > > HANGS till end of sleep
> > > > > > =============
> > > > > >
> > > > > > ==== Maxima ====
> > > > > > > :lisp (sleep 10)
> > > > > > PRESS QUIT
> > > > > > > 1
> > > > > > HANGS till end of sleep
> > > > > > ================
> > > > > >
> > > > > > Maybe also Maxima plugin is not catching the SIGINT...
> > > > > > In my plugin managed to receive the signal but I don't see it coming when i 
> > > > > > press the STOP button.
> > > > > >
> > > > > > Do you know something about it ?
> > > > > >
> > > > > > bye
> > > > > > Nicola

--- src/Plugins/Qt/qt_pipe_link.cpp.orig	2015-03-05 16:34:14 UTC
+++ src/Plugins/Qt/qt_pipe_link.cpp
@@ -9,6 +9,9 @@
 * in the root directory or <http://www.gnu.org/licenses/gpl-3.0.html>.
 ******************************************************************************/
 
+// for my debug
+#include<iostream>
+
 #include "basic.hpp"
 
 #if defined (QTTEXMACS) && (defined (__MINGW__) || defined (__MINGW32__) || defined (QTPIPES))
@@ -145,7 +148,19 @@ qt_pipe_link_rep::interrupt () {
 #if defined(__MINGW__) || defined(__MINGW32__)
   // Not implemented
 #else
-  ::killpg(PipeLink.pid (), SIGINT);
+  
+  // ::killpg(PipeLink.pid (), SIGINT);
+  
+  // for some reason the prevoius "kill" kills
+  // (-1)*myPid ! 
+  
+  Q_PID pid = PipeLink.pid();
+  cout << "DBG> signal SIGINT on pid: " << pid << "\n";  
+  int ret = ::kill(pid, SIGINT);
+  if (ret == -1) {
+     qt_error << "Interrupt not successful, pid: " << pid << " return code: " << errno << "\n";
+  }
+
 #endif
 }