The test suite for building SWI-Prolog is run by the command ctest, which uses the file packages/cmake/PrologPackage.cmake to run tests defined by add_test(NAME "${SWIPL_PKG}:${test_name}" COMMAND ${PROG_SWIPL} -p "foreign=${pforeign}" -f none --no-packs -s ${test_source} -g "${test_g…

I have also been bitten by this behavior but in a server environment with automatic code deployment. Code was deployed and started successfully despite some modules having loading errors. The errors would cause issues later when specific code path was run. I defined a dynamic flag “started” to coun…

Probably the best way out would be to provide some library that you can load at the start of your program or even using a commandline argument. As compiling and running are not well separated stages in SWI-Prolog there are many complicated cases though. For tests even more as there are tests tha…

It gets worse: a syntax error doesn’t result in an error from the compiler (swipl --stand_alone=true --undefined=error --verbose=false --foreign=save -o xxx.qlf -c xxx.pl). I hadn’t noticed this before because a syntax error usually cascades to other errors that do stop the compiler, Why is syntax …

Syntax error gives RC=0 from compiler opened 08:04PM - 03 May 21 UTC (UTC) [kamahen] kamahen The compiler should return a non-zero return code if there…

I’m still pretty unsure on how to address this. Yes, it is a problem in unattended builds. A robust solution that will not cause basically ok programs that print error messages to exit with status 1 is far from trivial. One option might be to do this only for swipl -c file.pl as that is not su…

[image] jan: One option might be to do this only for swipl -c file.pl This (partial) solution would at least produce “less surprise” when generating .qlf files in Makefiles. Perhaps controlled by a flag --error=error (to match --undefined=error)? I presume that if someone wants to suppress a …

Where in the source is the compiler called? (I tried tracing through the top-level intialization, but I got lost). Would I be better off by creating a user:message/3 hook, and then running qsave_program/2, rather than using swipl -c? But I’d need to somehow remove the user:message/3 hook from the s…

The -c invocation is handled by ‘$compile’/0 in boot/toplevel.pl. One solution to get rid of the hook is initialization/2 with When set to restore_state. Alternatively just before creating the state. Not sure anything can go wrong that doesn’t lead to failure anyway. Or create a generic handl…

That sounds like a non-trivial amount of work, so I guess we’ll just have to leave this as an open issue in github. For my purposes, it’s sufficient to have a user:message_hook/3 that I can abolish after loading the saved state - and it appears that user:message_hook/3 has global effect, so it seem…

It seems that there’s a solution, and a rather elegant one at that. Thank-you @jan ! [image] Changes in error handling Announce As a followup to Automatic tests can succeed with syntax errors and similar topics raised in the past I’ve pushed a number of changes: Add glob…

Automatic tests can succeed with syntax errors

Discussion

peter.ludemann May 8, 2021, 7:14pm 11

It seems that there’s a solution, and a rather elegant one at that. Thank-you @jan!

Topic		Replies	Views
Changes in error handling Announce	0	458	May 8, 2021
Disable warning message "% Disabled autoloading" General how-to	5	422	December 12, 2022
Issue building SWI-Prolog 8.5.14-17-g851d08ab6 Installing	2	414	July 28, 2022
Halt is giving ERROR: Prolog initialisation failed: with swipl version 9.3.12-20-g5d8b8d94e General	3	72	October 3, 2024
Strange error with make/0 Help!	4	593	May 30, 2019

Automatic tests can succeed with syntax errors

Related topics