Approach for adding more coverage in qofinstance.c

[John Ralls]

Yeah, that was a poor design on my part. Better to have two cases, the trivial one with 0 and another with some other number > 2. And it can lose the utterly pointless tests that list insertions increase the list length.

Regards,
John Ralls

> On Jan 22, 2026, at 05:44, Stefan Koch <stefan.koch.micro at gmail.com> wrote:
> 
> Thank you.  
> 
> I think the g_test_init is necessary mainly for the g_test_log_set_fatal_handler and/or the g_log_set_handler. So, I think I will need it even after I replace the glib random numbers.  
> 
> I personally don't like random numbers inside a test anyway.  Does it make sense to use a fixed number in the range instead?  
> 
> 
> On Thu, Jan 22, 2026 at 12:15 AM John Ralls <jralls at ceridwen.us <mailto:jralls at ceridwen.us>> wrote:
>> Stefan,
>> 
>> You write that you had to use TEST_F as if it was a bad thing. Using fixtures is a *good* thing. 
>> I think that the message handling is separate from g_test and doesn’t require g_test_init, but your custom main() also inits QOF and the logging system so I guess you’re stuck with it.
>> 
>> I made some comments on your commit. 
>> 
>> Regards,
>> John Ralls
>> 
>> 
>>> On Jan 21, 2026, at 11:53, Stefan Koch <stefan.koch.micro at gmail.com <mailto:stefan.koch.micro at gmail.com>> wrote:
>>> 
>>> (Thank you for merging my first commit test-qofid.cpp).
>>> 
>>> I am moving on to adding more coverage, in particular I thought in the qofinstance would be a good low level next step.
>>> 
>>> This is currently in part of the glib based test for test-oqf.  I am proposing removing it from there and making it a stand alone google test.  I like the stand alone test because it allows me just to run this one and see the coverage of just this test. (That removes other tests from providing accidental coverage.)
>>> 
>>> I also like full coverage since it is easy to check for 100% coverage, but hard to check if the coverage moved from 88% to 85% when running tests.  I cannot get 100% coverage since there are some lines that are unreachable when the other subsystems are correct.  I thought I saw some setup in lcov that allows marking of lines to not be covered, but I have not investigated that further.
>>> 
>>> There were some issues that I had to work out:
>>> 1. I had to use the google TEST_F so that I could have a class to replace the setup and teardown from glib testing.
>>> 2. I had to create a main() that overrides the default google one, that allows the setting g_test_init which is still needed since the qofinstance still makes use of the glib error handling.
>>> 3. I replaced the g_assert functions with the equivalent google test calls.
>>> 4. I replaced the g_test_message with a local call that does printf.
>>> 
>>> The I have not added extra coverage.  I wanted to verify that this approach is what is desired here, or if this is too to change, and not worth your time reviewing.
>>> 
>>> The commit can be seen here: https://github.com/stefan-koch-micro/gnucash/commit/fc5742e12d8007f86929450080750a14bec3ec43#diff-dcd2eece63e17b3ec55c96a3eab8a60c4582884a2d01b47b1c4940c1c0594bf0
>>> I have not made a pull request as this is a partial implementation. If the approach is OK, I will add more/full coverage and then make a merge request. (Or even do more of the test-qof set before making the request.)
>>> _______________________________________________
>>> gnucash-devel mailing list
>>> gnucash-devel at gnucash.org <mailto:gnucash-devel at gnucash.org>
>>> https://lists.gnucash.org/mailman/listinfo/gnucash-devel
>> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gnucash.org/pipermail/gnucash-devel/attachments/20260122/9bbdc512/attachment.htm>