> This bug clearly shows the limits of the component-based testing approach
with runtests. I think we shoud consider some user-level testing tool which
includes > GUI record/replay. maybe based on
http://drdobbs.com/tools/184404691 .
http://web.archive.org/web/20040604025740/http://www.wildopensource.com/larry-projects/android.html
enjoy.
j.
On Fri, Nov 18, 2011 at 10:05 AM, Michael Haberler <mai...@mah.priv.at>wrote:
> every tried to exit from an oword subroutine with an M2 and wondered what
> happened?
>
> this is what happens:
> - assume interpreter is executing a named oword sub in a separate file
> - when the interpreter switches from the main ngc file to the sub, it
> changes the filename to the sub's filename
> - the interpreter executes M2 and does NOT return from the sub, leaving
> the open filename the sub's filename
> - running the program again now executes the sub ngc file, with
> potentially undesired code execution
>
> To exercise, an Axis config is needed:
> - fetch
> http://git.mah.priv.at/gitweb/emc2-dev.git/shortlog/refs/heads/fix-m2-oword
> - pick 515be6
> - cd to configs/sim/bug and execute 'emc bug.ini'
> - hit the run button twice
>
> For the fix, pick the second commit fb8698e and build; it seems m2 can now
> safely be used to terminate program execution from a sub
>
> This is basically the same horror bug as the one below (fixed in
> 009b76630), just triggered from a differnt corner of the interpreter (note
> the message created by '(debug, should not happen!)' in fail.ngc on the
> second execution:
>
>
> http://git.mah.priv.at/gitweb/emc2-dev.git/shortlog/refs/heads/failed-oword-regression-test
>
> http://git.mah.priv.at/gitweb/emc2-dev.git/shortlog/refs/heads/fix-failed-oword-procedure
>
> Howeber, this bug cannot be tested for with a rs274-based regression test
> because the state must be viewed with a debugger or within a complete gui
> based application - rs274 exits before it gets interesting
>
> Actually I suspect bug 1734309 to be related
>
> since this can cause accidential code execution I'll push this one right
> away to master; I'll look into redoing this with v2.5_branch which needs a
> different approach
>
> -----
>
> This bug clearly shows the limits of the component-based testing approach
> with runtests. I think we shoud consider some user-level testing tool which
> includes GUI record/replay. maybe based on
> http://drdobbs.com/tools/184404691 .
>
> -Michael
>
> ------------------------------------------------------------------------------
> All the data continuously generated in your IT infrastructure
> contains a definitive record of customers, application performance,
> security threats, fraudulent activity, and more. Splunk takes this
> data and makes sense of it. IT sense. And common sense.
> http://p.sf.net/sfu/splunk-novd2d
> _______________________________________________
> Emc-developers mailing list
> Emc-developers@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-developers
>
------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure
contains a definitive record of customers, application performance,
security threats, fraudulent activity, and more. Splunk takes this
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers
