Exact named qct/0036-notneg.c source compiles to a main function returning zero on MCS51 and DS390; generated return instructions were inspected. Runtime execution was unnecessary for this constant return, and was not performed. https://godbolt.org/z/jMbhqKvdd
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
@maartenbrock Sorry for the trouble. I chose to do that because I wanted to follow a consistent methodology, and I did not know at that time how/whether the existing test is run, or how it manifests success or failure. I did not think that regression test time would be significantly affected by such a simple test. I also thought it would be easier to find a regression test named after the bug, for auditing purposes.
I think that simply deleting sdcc/support/regression/tests/bug-3392.c might leave a coverage gap because of an exclusion renewed in [r16905].
I can still reproduce the bug using current [r14409].
Related
Commit: [r14409]
I found that the bug has been fixed in [r16281], but I don't know which part of the code logic corresponds to it, and I want to know.
Related
Commit: [r16281]
Last edit: Maarten Brock 2026-10-20
Exact named qct/0036-notneg.c source compiles to a main function returning zero on MCS51 and DS390; generated return instructions were inspected. Runtime execution was unnecessary for this constant return, and was not performed.
https://godbolt.org/z/jMbhqKvdd
@spth fixed this in [r15572] along with bug [#3355].
Related
Bugs:
#3355Commit: [r15572]
Last edit: Maarten Brock 2026-10-20
Added regression test in [r17058]
Related
Commit: [r17058]
Why? It was already tested by qct/0036-notneg.c.
SDCC does not need even longer regression test time.
@maartenbrock Sorry for the trouble. I chose to do that because I wanted to follow a consistent methodology, and I did not know at that time how/whether the existing test is run, or how it manifests success or failure. I did not think that regression test time would be significantly affected by such a simple test. I also thought it would be easier to find a regression test named after the bug, for auditing purposes.
I think that simply deleting sdcc/support/regression/tests/bug-3392.c might leave a coverage gap because of an exclusion renewed in [r16905].
Related
Commit: [r16905]
[r17069] and [r17070] enables the original regression test for bug #3392 on MCS51 and DS390 and deletes its duplicate, tests/bug-3392.c.
Related
Commit: [r17069]
Commit: [r17070]
Last edit: Christopher Bazley 2026-10-22
Thanks, that is exactly how it should be IMHO.
And good catch that I forgot to reenable the original test.