Fix bugs in dialyzer task #49

UncleGrumpy · 2025-06-18T04:29:14Z

Fixes several bugs in the atomvm dialyzer task. Beam paths are correctly discovered when using the AtomVM/build directory as the atomvm_root. A bug which caused the plt check to always fail, causing a complete rebuild was fixed. The full atomvm version is used in the base PLT name so PLTs from different branches can be used without loosing the previous ones. The task will now abort when an error occurs rather than try to keep going. Improved error messages, stack traces are only printed (with better formatting) when running rebar3 with DEBUG=1.

Fixes a bug when checking the base PLT which caused it to always be rebuilt. Signed-off-by: Winford <[email protected]>

Keep the full version when naming the base PLT file to allow greating base PLTs for differnt git commits (or different branches) when working on the same base version. Signed-off-by: Winford <[email protected]>

pguyot

Unrelated to these changes, there is a problem with the paths.
I needed to comment or move these lines:
rebar_paths:unset_paths([plugins], State),
rebar_paths:set_paths([deps], State),

Also, with atomvm_benchmark project, when building the app PLT I got a lot of undefined functions that should be in AtomVM PLT.

benchmark.erl:0: Unknown function erlang:get_module_info/1
benchmark.erl:0: Unknown function erlang:get_module_info/2
benchmark.erl:15:24: Unknown function atomvm:platform/0
benchmark.erl:19:21: Unknown function io:format/2

And subsequent calls work, but I have a failure again if I delete the benchmark.plt file.

src/atomvm_dialyzer_provider.erl

UncleGrumpy · 2025-06-18T22:24:26Z

Unrelated to these changes, there is a problem with the paths. I needed to comment or move these lines: rebar_paths:unset_paths([plugins], State), rebar_paths:set_paths([deps], State),

I forgot to investigate those closer, it turns out they should not have been used at all. I removed them.

Also, with atomvm_benchmark project, when building the app PLT I got a lot of undefined functions that should be in AtomVM PLT.
benchmark.erl:0: Unknown function erlang:get_module_info/1
benchmark.erl:0: Unknown function erlang:get_module_info/2
benchmark.erl:15:24: Unknown function atomvm:platform/0
benchmark.erl:19:21: Unknown function io:format/2
And subsequent calls work, but I have a failure again if I delete the benchmark.plt file.

I will look into this and see what I did wrong.

UncleGrumpy · 2025-06-18T23:57:56Z

It turned out that I just needed to ignore warnings when generating the app PLT. The app PLT was being created properly but also issuing warnings when being built that should be ignored (as long as they aren't errors) until the actual analysis is done after making sure the base and application PLT are generated or checked if they are out of date.

pguyot · 2025-06-19T04:19:59Z

With the latest commit (ada37af), I get the following warning

     ┌─ _build/default/plugins/atomvm_rebar3_plugin/src/atomvm_dialyzer_provider.erl:
     │
 222 │  do_build_plt(Config, State) ->
     │               ╰── Warning: variable 'Config' is unused

Other than that, a quick test shows it behaves as expected 💯
For example:

it relates if I use a function that is not declared in AtomVM libs but exists with OTP (I tested with io:rows/0).
it generates an error if I use a function that that does exist in either AtomVM libs or OTP (I tested with io:formatz/2)
it detects type error within my module (I tried to match with a tuple a value in a function that was the result of either erlang:system_time(microsecond) or erlang:system_time(millisecond) * 1000).
it detects contract errors for NIFs (it detects a call to erlang:system_time(millisecondz)

It does not detect some items related to types that we should add. For example it didn't detect a call to erlang:system_flag(schedulers_onlinez, 1), but I guess that's because we don't define the flags that are valid in the spec (at least in the AtomVM version that is installed on my machine).

Thank you! It shoud prove a very valuable tool.

UncleGrumpy · 2025-06-19T05:48:38Z

With the latest commit (ada37af), I get the following warning

     ┌─ _build/default/plugins/atomvm_rebar3_plugin/src/atomvm_dialyzer_provider.erl:
     │
 222 │  do_build_plt(Config, State) ->
     │               ╰── Warning: variable 'Config' is unused

Thank you I missed that at the beginning since I ran in diagnostic mode the last few times and didn't scroll back far enough when I got the results I was looking for. Fixed now.

pguyot · 2025-06-19T05:52:04Z

I confirm the warning is gone with 076adff

pguyot · 2025-06-19T05:55:43Z

src/atomvm_dialyzer_provider.erl

            {error, no_beams_found};
        Ebins ->
            Ebins
    end.

+% @private
+libs_not_found() ->


This looks like lists:duplicate/2 to me.

pguyot · 2025-06-19T05:57:57Z

src/atomvm_dialyzer_provider.erl

    Libs =
        case [filelib:wildcard(Base ++ "/lib/" ++ atom_to_list(Lib) ++ "*/ebin") || Lib <- ?LIBS] of
-            [] ->
-                [filename:join(Base, atom_to_list(Lib2) ++ "/src/beams") || Lib2 <- ?LIBS];
+            NotFound ->


This logic could be simplified or improved.
Do we want to add for each lib in ?LIBS either ebin or src/beams?

I would suggest doing a foldl on the list instead than matching with [[], [], ..., []].

Fixes a bug where the beam files are not found when using source builds from Atomvm/build directory as the atomvm_root path. Signed-off-by: Winford <[email protected]>

Removes unnecessary path changes in the do/1. Signed-off-by: Winford <[email protected]>

Fixes several instances where the task would try to complete after encountering an error. Improves output when errors do occur by printing an informative message under noral opperation, and only print the full stacktrace when DEBUG=1 is used. When displayed stacktraces are formatted better by starting them on a new line. Signed-off-by: Winford <[email protected]>

Removes unnecessary checks for warnings when building the application PLT for the first time, these should be ignored until the actual analysis after both the base and application PLT are checked or created. Signed-off-by: Winford <[email protected]>

UncleGrumpy added 2 commits June 17, 2025 21:20

Fix bug in plt check

f1468f1

Fixes a bug when checking the base PLT which caused it to always be rebuilt. Signed-off-by: Winford <[email protected]>

Stop truncating AtomVM version

57f42fa

Keep the full version when naming the base PLT file to allow greating base PLTs for differnt git commits (or different branches) when working on the same base version. Signed-off-by: Winford <[email protected]>

UncleGrumpy force-pushed the fix_dialyzer_bugs branch 2 times, most recently from 54b4fb1 to 8bf9af3 Compare June 18, 2025 04:43

UncleGrumpy requested a review from pguyot June 18, 2025 05:28

pguyot reviewed Jun 18, 2025

View reviewed changes

UncleGrumpy force-pushed the fix_dialyzer_bugs branch from 8bf9af3 to 8d933ce Compare June 18, 2025 21:44

UncleGrumpy force-pushed the fix_dialyzer_bugs branch 3 times, most recently from ea789d2 to ada37af Compare June 18, 2025 23:54

UncleGrumpy force-pushed the fix_dialyzer_bugs branch from ada37af to 076adff Compare June 19, 2025 05:35

pguyot reviewed Jun 19, 2025

View reviewed changes

UncleGrumpy added 4 commits June 19, 2025 02:40

Fix bug when using AtomVM/build as source beams for building PLT

86c235f

Fixes a bug where the beam files are not found when using source builds from Atomvm/build directory as the atomvm_root path. Signed-off-by: Winford <[email protected]>

Don't change rebar paths

e4f9541

Removes unnecessary path changes in the do/1. Signed-off-by: Winford <[email protected]>

UncleGrumpy force-pushed the fix_dialyzer_bugs branch from 076adff to 98895fa Compare June 19, 2025 09:41

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Fix bugs in dialyzer task #49

Fix bugs in dialyzer task #49

Uh oh!

UncleGrumpy commented Jun 18, 2025

Uh oh!

pguyot left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

UncleGrumpy commented Jun 18, 2025

Uh oh!

UncleGrumpy commented Jun 18, 2025

Uh oh!

pguyot commented Jun 19, 2025

Uh oh!

UncleGrumpy commented Jun 19, 2025

Uh oh!

pguyot commented Jun 19, 2025

Uh oh!

pguyot Jun 19, 2025

Uh oh!

pguyot Jun 19, 2025

Uh oh!

Uh oh!

Fix bugs in dialyzer task #49

Are you sure you want to change the base?

Fix bugs in dialyzer task #49

Uh oh!

Conversation

UncleGrumpy commented Jun 18, 2025

Uh oh!

pguyot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

UncleGrumpy commented Jun 18, 2025

Uh oh!

UncleGrumpy commented Jun 18, 2025

Uh oh!

pguyot commented Jun 19, 2025

Uh oh!

UncleGrumpy commented Jun 19, 2025

Uh oh!

pguyot commented Jun 19, 2025

Uh oh!

pguyot Jun 19, 2025

Choose a reason for hiding this comment

Uh oh!

pguyot Jun 19, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!