On Tue, 21 Sep 2021, Sven Schreiber wrote: Well, I'm not stopping you! ;) ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------

Hi Hélio, Good idea. By the way, we have in our workspace repository a folder where to put tests into. Have a look here, please: https://sourceforge.net/p/gretl/workspace/ci/master/tree/tests/README.md I think it would be nice to write this test using gretl and/ or bash. The process could be: 1. Create a ascii file with all commands parsed from somewhere (I think we already a have a job for this); delimited by \n 2. Execute a gretl job which a) loops over each line of this ascii file, b) store a text file (foo.inp) with the content: 'CMD' where 'CDM' is for instance 'ols', c) execute another gretl job using the 'shell' command: 'gretlcli foo.inp' d) Retrieve the exit code of this process, e) Finish the loop, f) List all commands which let to some unusual error code (referring to a segfault or so). Best, Artur Am 21.09.2021 15:50 schrieb Hélio Guilherme: > Those kind of tests should be easily automated. > > Can some one write Python code for this pseudo-code? > > Get a list of all commands (from the code) > Iterate all commands without arguments (what to expect?) > > Later this can be a test triggered on each code pushed to master. > > On Tue, Sep 21, 2021 at 2:43 PM Riccardo (Jack) Lucchetti > <p002264(a)staff.univpm.it&gt; wrote: > >> On Tue, 21 Sep 2021, Sven Schreiber wrote: >> >>> Am 21.09.2021 um 14:51 schrieb Riccardo (Jack) Lucchetti: >>>> >>>> The "sscanf" command is deprecated, so this is probably not very >>>> important, however: I just discovered a way to trigger a >> segfault. >>>> Just open a console or gretlcli, type "sscanf" and hit Enter. >> Boom. >>>> >>>> Fortunately, the fix was easy and I just pushed it to git. >>>> >>> Thanks, Jack! >>> >>> Maybe it would be a good idea to trivially test all the commands >> that we >>> have (or have had) by just typing them without any parameters, >> like you >>> did with sscanf !? >> >> Well, I'm not stopping you! ;) >> >> ------------------------------------------------------- >> Riccardo (Jack) Lucchetti >> Dipartimento di Scienze Economiche e Sociali (DiSES) >> >> Università Politecnica delle Marche >> (formerly known as Università di Ancona) >> >> r.lucchetti(a)univpm.it >> http://www2.econ.univpm.it/servizi/hpp/lucchetti [1] >> > -------------------------------------------------------_______________________________________________ >> Gretl-devel mailing list -- gretl-devel(a)gretlml.univpm.it >> To unsubscribe send an email to gretl-devel-leave(a)gretlml.univpm.it >> Website: >> > https://gretlml.univpm.it/postorius/lists/gretl-devel.gretlml.univpm.it/ >> [2] > > > Links: > ------ > [1] http://www2.econ.univpm.it/servizi/hpp/lucchetti > [2] > https://gretlml.univpm.it/postorius/lists/gretl-devel.gretlml.univpm.it/ > > _______________________________________________ > Gretl-devel mailing list -- gretl-devel(a)gretlml.univpm.it > To unsubscribe send an email to gretl-devel-leave(a)gretlml.univpm.it > Website: > https://gretlml.univpm.it/postorius/lists/gretl-devel.gretlml.univpm.it/

Am 21.09.2021 um 16:14 schrieb Sven Schreiber: Sorry, to be useful the crucial line should include "catch", like this: <hansl> loop foreach i sscanf    # add more command names later     catch $i endloop </hansl> Still crashes now, but shouldn't after the bug is fixed. cheers sven

Am 21.09.2021 17:55 schrieb Hélio Guilherme: Hi Hélio, Currently we have the source code of our workspace on sourceforge. Even though personally I also favor github, we need to discuss whether we want to move the test-suit (I only mean: https://sourceforge.net/p/gretl/workspace/ci/master/tree/tests) to the gretl github repo: https://github.com/gretl-project I could create a new repo named "gretl-tests" or so and move /master/tree/tests if _all_ agree. The nice about github is automatic or scheduled testing: Via github-action or so, we could schedule some container (e.g. Ubuntu) which starts once a day, pulls the latest gretl code, runs the test suit and sends an email in case of failures. Best, Artur > On Tue, Sep 21, 2021 at 3:17 PM Sven Schreiber <svetosch(a)gmx.net&gt; > wrote: > >> Am 21.09.2021 um 16:14 schrieb Sven Schreiber: >>> But actually, a pure hansl script can go a long way I believe. >> Like this: >>> <hansl> >>> loop foreach i sscanf # add more command names later >>> $i >>> endloop >>> </hansl> >>> >>> This already crashes the GUI program for me here. >> >> Sorry, to be useful the crucial line should include "catch", like >> this: >> >> <hansl> >> >> loop foreach i sscanf # add more command names later >> catch $i >> endloop >> >> </hansl> >> >> Still crashes now, but shouldn't after the bug is fixed. >> >> cheers >> >> sven >> _______________________________________________ >> Gretl-devel mailing list -- gretl-devel(a)gretlml.univpm.it >> To unsubscribe send an email to gretl-devel-leave(a)gretlml.univpm.it >> Website: >> > https://gretlml.univpm.it/postorius/lists/gretl-devel.gretlml.univpm.it/ >> [1] > > > Links: > ------ > [1] > https://gretlml.univpm.it/postorius/lists/gretl-devel.gretlml.univpm.it/ > [2] https://github.com/HelioGuilherme66/gretl > > _______________________________________________ > Gretl-devel mailing list -- gretl-devel(a)gretlml.univpm.it > To unsubscribe send an email to gretl-devel-leave(a)gretlml.univpm.it > Website: > https://gretlml.univpm.it/postorius/lists/gretl-devel.gretlml.univpm.it/

On Tue, 21 Sep 2021, Riccardo (Jack) Lucchetti wrote: Testing is always good, but I don't think this is something we need get too excited about. The bug that Jack found was in an exceptional case: a redirect from a defunct command to a function. For commands that are in good standing we have guards in place against missing arguments. OK, I can't guarantee that the guards are 100% bullet-proof, but I'd say better than 99%. Allin