7.5: Internal error after enabling horizontal or vertical graph

[kglotfelty]

#30 Issue by kglotfelty,

#30 Issue by olebole,

When I enable one of the graphs in the "View" menu, I get an error box "An internal error has been detected [OK]" as soon as the mouse is in the image area, but outside of the borders of the FITS image.
This window immediately re-appears as long as I don't move it to a place where the cursor is not left over such an area after the error box disappears.
On the Debian build, I also get a segmentation fault when I am in within the borders of a FITS image, when calling Base::bltCut(char*, char*, Coord::Orientation, Vector const&) () in frame/frblt.C line 947.
This does not appear in the original built. The major difference here are that Debian uses specific build flags to enable hardening (-fstack-protector-strong -D_FORTIFY_SOURCE=2) and the we use tkblt as a separate tcl package. I will also investigate the crash myself, but it may help with debugging the other problem.

[kglotfelty]

Comment by kglotfelty:

Comment by wjoye:

Ole

Thanks for the note. The graph problem has already been fixed. I think the 2 issues are the same problem. In ds9 I set up a trap for signals and try to gracefully recover. I'm on vacation right now and will not be back until late next week.

Regards

Bill

Sent from my iPad

On Apr 7, 2017, at 3:57 AM, Ole Streicher notifications@github.com wrote:

When I enable one of the graphs in the "View" menu, I get an error box "An internal error has been detected [OK]" as soon as the mouse is in the image area, but outside of the borders of the FITS image.
This window immediately re-appears as long as I don't move it to a place where the cursor is not left over such an area after the error box disappears.
On the Debian build, I also get a segmentation fault when I am in within the borders of a FITS image, when calling Base::bltCut(char*, char*, Coord::Orientation, Vector const&) () in frame/frblt.C line 947.
This does not appear in the original built. The major difference here are that Debian uses specific build flags to enable hardening (-fstack-protector-strong -D_FORTIFY_SOURCE=2) and the we use tkblt as a separate tcl package. I will also investigate the crash myself, but it may help with debugging the other problem.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.

[kglotfelty]

Comment by kglotfelty:

Comment by olebole:

That is great! Once you are back, could I ask you to point me to the fix? I want to get it into Debian Stretch, but since we are on a freeze, no new versions are accepted, but just well-defined fixes.

[kglotfelty]

Comment by kglotfelty:

Comment by wjoye:

Internal error fixed previously. Unable to reproduce segv.

[kglotfelty]

Comment by kglotfelty:

Comment by olebole:

The "internal error" is fixed with edea206, but after applying, the segfault still appears in the Debian package. I just opened a Debian bug (#859777) for this (needed anyway to get a new version in before release) and investigate it. If I find something to fix in the source, I will get back here.

[kglotfelty]

Comment by kglotfelty:

Comment by olebole:

I found the problem: In Debian, we use the tkblt package dynamically (with stubs), and I forgot to initialize the stubs. Therefore, all function pointers into tkblt were uninitialized. I am wondering now why this worked ever...
The patch that I applied is however here.