GLib-CRITICAL errors during build of docs (dev octave)

[PhilipNienhuis]

When building dev-octave, during generation of docs, I get since (I
think) yesterday or day before yesterday, many Glib-CRITICAL errors
along the lines of:

:
   GEN      doc/interpreter/delaunay.txt

(process:28681): GLib-CRITICAL **: g_source_unref_internal: assertion
'source != NULL' failed
:

Nevertheless the build continues fine. I suppose all bets are off for
the figures in the generated docs.

Anyone else seeing this as well?

(This is on both my build systems, Mageia-5 and Mageia-6 linux)

Philip

[Dmitri A. Sergatskov]

On Fri, Sep 1, 2017 at 3:50 PM, Philip Nienhuis <[hidden email]> wrote:

When building dev-octave, during generation of docs, I get since (I think) yesterday or day before yesterday, many Glib-CRITICAL errors along the lines of:

:
  GEN      doc/interpreter/delaunay.txt

(process:28681): GLib-CRITICAL **: g_source_unref_internal: assertion 'source != NULL' failed
:

Nevertheless the build continues fine. I suppose all bets are off for the figures in the generated docs.

Anyone else seeing this as well?

(This is on both my build systems, Mageia-5 and Mageia-6 linux)

Philip

​I saw this, then I re-cloned and it went away.​

​Dmitri.

--

​

[PhilipNienhuis]

Dmitri A. Sergatskov wrote:

>
>
> On Fri, Sep 1, 2017 at 3:50 PM, Philip Nienhuis <[hidden email]
> <mailto:[hidden email]>> wrote:
>
>     When building dev-octave, during generation of docs, I get since (I
>     think) yesterday or day before yesterday, many Glib-CRITICAL errors
>     along the lines of:
>
>     :
>       GEN      doc/interpreter/delaunay.txt
>
>     (process:28681): GLib-CRITICAL **: g_source_unref_internal:
>     assertion 'source != NULL' failed
>     :
>
>     Nevertheless the build continues fine. I suppose all bets are off
>     for the figures in the generated docs.
>
>     Anyone else seeing this as well?
>
>     (This is on both my build systems, Mageia-5 and Mageia-6 linux)
>
>     Philip
>
>
> ​I saw this, then I re-cloned and it went away.​

One of my systems has been freshly installed, so I did have a fresh
clone there as well...

Philip

[Dmitri A. Sergatskov]

On Fri, Sep 1, 2017 at 4:40 PM, Philip Nienhuis <[hidden email]> wrote:

Dmitri A. Sergatskov wrote:

On Fri, Sep 1, 2017 at 3:50 PM, Philip Nienhuis <[hidden email]
<mailto:[hidden email]>> wrote:

    When building dev-octave, during generation of docs, I get since (I
    think) yesterday or day before yesterday, many Glib-CRITICAL errors
    along the lines of:

    :
      GEN      doc/interpreter/delaunay.txt

    (process:28681): GLib-CRITICAL **: g_source_unref_internal:
    assertion 'source != NULL' failed
    :

    Nevertheless the build continues fine. I suppose all bets are off
    for the figures in the generated docs.

    Anyone else seeing this as well?

    (This is on both my build systems, Mageia-5 and Mageia-6 linux)

    Philip


​I saw this, then I re-cloned and it went away.​

One of my systems has been freshly installed, so I did have a fresh clone there as well...

​Sorry , -- I do see them. ​ My error though:

(lt-octave-gui:2306): GLib-CRITICAL **: g_source_unref_internal: assertion 'source != NULL' failed

 

Philip

​Dmitri.

​

[Mike Miller-4]

On Fri, Sep 01, 2017 at 17:15:54 -0500, Dmitri A. Sergatskov wrote:
> (lt-octave-gui:2306): GLib-CRITICAL **: g_source_unref_internal: assertion 'source != NULL' failed

I think this is caused when multiple instances of Octave are generating
the plot figures concurrently, with make -jN. I can get these warnings
to go away with make -j1.

I can make this message appear sometimes by just running a few of the
plot scripts in parallel from the shell, for example

  $ ./run-octave --path ../doc/interpreter \
                 --eval 'sparseimages(pwd, "spchol", "pdf")' & \
    ./run-octave --path ../doc/interpreter \
                 --eval 'sparseimages(pwd, "spcholperm", "pdf")' & \
    ./run-octave --path ../doc/interpreter \
                 --eval 'sparseimages(pwd, "spmatrix", "pdf")' &

Running that a few times is enough to trigger the message for me.

--
mike

[Mike Miller-4]

On Fri, Sep 01, 2017 at 16:00:11 -0700, Mike Miller wrote:
> I think this is caused when multiple instances of Octave are generating
> the plot figures concurrently, with make -jN. I can get these warnings
> to go away with make -j1.

Backing out changeset fb9b024a6041 makes these messages go away. But
that change was meant to fix bug #50210. Did this change introduce
another problem or are these messages safe to ignore? Does anyone know?

--
mike

[Dmitri A. Sergatskov]

On Fri, Sep 1, 2017 at 6:16 PM, Mike Miller <[hidden email]> wrote:

On Fri, Sep 01, 2017 at 16:00:11 -0700, Mike Miller wrote:
> I think this is caused when multiple instances of Octave are generating
> the plot figures concurrently, with make -jN. I can get these warnings
> to go away with make -j1.

Backing out changeset fb9b024a6041 makes these messages go away. But
that change was meant to fix bug #50210. Did this change introduce
another problem or are these messages safe to ignore? Does anyone know?

​I rebuilt docs with -j8 and -j1 and I do not see any difference.​

 

--
mike

​Dmitri.

--

​

[Mike Miller-4]

On Fri, Sep 01, 2017 at 19:16:24 -0500, Dmitri A. Sergatskov wrote:
> I rebuilt docs with -j8 and -j1 and I do not see any difference.

When my system is heavily loaded, I can get Octave to issue this message
with something as simple as

    $ octave --eval true

--
mike

[Daniel Sebald]

On 09/01/2017 06:16 PM, Mike Miller wrote:
> On Fri, Sep 01, 2017 at 16:00:11 -0700, Mike Miller wrote:
>> I think this is caused when multiple instances of Octave are generating
>> the plot figures concurrently, with make -jN. I can get these warnings
>> to go away with make -j1.
>
> Backing out changeset fb9b024a6041 makes these messages go away. But
> that change was meant to fix bug #50210. Did this change introduce
> another problem or are these messages safe to ignore? Does anyone know?

Yes, I think it is creating another problem.  The changeset looks like
something I would have tried initially when attempting a robust exit
sequence when the user selects "Exit" from the main GUI window while
Octave-core is busy processing.  But terminate() and wait() don't work
quite the way one would hope.  terminate() isn't graceful stoppage of
the thread, instead it issues an error just as when Octave CLI were
terminated.

The previous work is

https://savannah.gnu.org/bugs/?44485

In fact, Philip is the person I was working with on testing that, but
unfortunately I never felt I handled all the various scenarios that
could occur, e.g., user pressing exit several times in row, or
Octave-core finally acknowledging it is complete while the "Should I
force termination?" dialog box is on the screen.  (The dialog appeared
after a two second timer expired.)

As I recall, my thinking at the time was to wait until Cntrl-C handling
was all worked out because that would be the proper way to stop Octave
if the user confirms "Yes, force termination" in the dialog box.  You'll
probably want to backout this change, as I don't think it can be made
robust.

For the documentation, I suppose using octave-cli as opposed to the GUI
isn't possible because graphics_toolkit qt is needed:

sebald@ ~ $ octave-cli
GNU Octave, version 4.3.0+
[snip]
octave:1> graphics_toolkit qt
error: graphics_toolkit: qt toolkit is not available
error: called from
     graphics_toolkit at line 81 column 5
octave:1> graphics_toolkit fltk

Dan

[Dmitri A. Sergatskov]

On Fri, Sep 1, 2017 at 8:35 PM, Mike Miller <[hidden email]> wrote:

On Fri, Sep 01, 2017 at 19:16:24 -0500, Dmitri A. Sergatskov wrote:
> I rebuilt docs with -j8 and -j1 and I do not see any difference.

When my system is heavily loaded, I can get Octave to issue this message
with something as simple as

​I saw warnings with make -j8 and did not see them with make -j1.

I meant that I do not see difference in the resulting docs.​

 

--
mike

​Dmitri​

[Dmitri A. Sergatskov]

​May be just octave trying to be feature-complete:

​

https://www.mathworks.com/matlabcentral/answers/250821-i-got-glib-warning-and-glib-critical

​Dmitri.

--

​

[John W. Eaton]

On 09/01/2017 09:35 PM, Mike Miller wrote:
> On Fri, Sep 01, 2017 at 19:16:24 -0500, Dmitri A. Sergatskov wrote:
>> I rebuilt docs with -j8 and -j1 and I do not see any difference.
>
> When my system is heavily loaded, I can get Octave to issue this message
> with something as simple as
>
>      $ octave --eval true

I pushed this changeset:

   http://hg.savannah.gnu.org/hgweb/octave/rev/8e4abfaadd97

Does it fix the problem you were seeing?

jwe

[Dmitri A. Sergatskov]

On Mon, Sep 4, 2017 at 9:46 PM, John W. Eaton <[hidden email]> wrote:

On 09/01/2017 09:35 PM, Mike Miller wrote:

On Fri, Sep 01, 2017 at 19:16:24 -0500, Dmitri A. Sergatskov wrote:

I rebuilt docs with -j8 and -j1 and I do not see any difference.

When my system is heavily loaded, I can get Octave to issue this message
with something as simple as

     $ octave --eval true

I pushed this changeset:

  http://hg.savannah.gnu.org/hgweb/octave/rev/8e4abfaadd97

Does it fix the problem you were seeing?

​That seems to fix the problem for me.​

 

jwe

​Dmitri.

--

​