www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - [SAOC 2024] Improve D Error Messages - Weekly Update #5

reply Royal Simpson Pinto <royalpinto007 gmail.com> writes:


This week, I focused on addressing feedback and adding tests for 
SARIF (Static Analysis Results Interchange Format) support in the 
DMD compiler. The current PR introduces a `--sarif` compiler flag 
to generate structured SARIF output for error messages.

**PR:** [feat(errors): add SARIF report generation 
https://github.com/dlang/dmd/pull/16986)



- **Adding Tests and Addressing Feedback**
   I spent time this week addressing the feedback provided on my 
PR, particularly refining the SARIF implementation. I added a 
test case to validate the output and ensure the correctness of 
SARIF reports.

- **Handling Regex Matching for Variable Outputs**
   I encountered issues with matching variable outputs in test 
pipelines versus local outputs. After multiple attempts to 
resolve these through regexes, I decided to implement a simpler 
version of the SARIF report, excluding certain complexities such 
as relative paths, to avoid complications.

- **Milestone 1 Report Preparation**
   I also dedicated time to finalizing and submitting my Milestone 
1 report, which summarizes the progress from September 15 to 
October 14, 2024.

- **Exploring Milestone 2 Tasks**
   I began exploring the objectives of Milestone 2, which involve 
enhancing error message utilities such as adding more detailed 
squiggles and error codes. I reviewed the existing utilities and 
documentation to understand how error codes and squiggles are 
currently implemented in the DMD compiler.



- **PR Merge**
   I had hoped my PR would be merged this week, which would allow 
me to start working on additional updates like using 
`MessageStyle` for SARIF output instead of a compiler flag. 
However, the merge is still pending, which is currently blocking 
me from making further progress on SARIF-related updates. 
Although the PR has been approved by one maintainer, it still 
needs to be merged.

- **Resolving Regex Issues**
   The most significant challenge was handling regexes to match 
variable outputs between local tests and CI pipelines. 
Ultimately, I simplified the SARIF version to avoid complexities 
related to relative paths and other variable outputs.



- **MessageStyle Implementation**
   Once the current PR is merged, I plan to work on implementing 
SARIF using `MessageStyle`, which will be done in a separate PR.

- **Further Exploration of Milestone 2**
   I will continue researching the error message utilities and 
explore how to integrate more detailed squiggles and error codes 
into the compiler. Any links to existing Bugzilla issues 
(specifically related to this) or resources would be really 
helpful.

---
Oct 21
parent reply Johan <j j.nl> writes:
On Monday, 21 October 2024 at 08:44:29 UTC, Royal Simpson Pinto 
wrote:



 This week, I focused on addressing feedback and adding tests 
 for SARIF (Static Analysis Results Interchange Format) support 
 in the DMD compiler. The current PR introduces a `--sarif` 
 compiler flag to generate structured SARIF output for error 
 messages.


Hi,
   It's nice that you are working on this!

I'd like to strongly urge you to put priority on these two things:
- Commandline interface. Please make this the same as GCC and 
Clang, such that tools do not have to implement a D-specific 
interface, and in may "just work" without any change needing in 
tooling. This includes having the option to output SARIF to a 
file.
- Output format. Mimic GCC and Clang as much as possible in what 
you output. Changes needed: (a) do not output errors to the same 
output stream as both normal and SARIF formats, because that 
immediately breaks tooling compatibility. (b) specify SARIF 
schema and adhere to it [1]. The output of your PR is not valid 
SARIF output [2]. For example, "invocation" is used wrong (should 
be inside a "runs" list).

The reason for this priority over anything else, is that both 
these are likely necessary to actually be able to use dlang+SARIF 
with external tools. So whatever other work you'll do will only 
be usable _after_ these two items are done.

cheers,
   Johan

[1] I again strongly advise to do what clang and GCC do, because 
it means that tooling will at least support that schema. They are 
using version 2.1.0 apparently. 
(https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json)

[2] https://sarifweb.azurewebsites.net/Validation
Oct 21
parent Royal Simpson Pinto <royalpinto007 gmail.com> writes:
On Monday, 21 October 2024 at 19:46:56 UTC, Johan wrote:


 [1] I again strongly advise to do what clang and GCC do, 
 because it means that tooling will at least support that 
 schema. They are using version 2.1.0 apparently. 
 (https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json)

 [2] https://sarifweb.azurewebsites.net/Validation


Hi Johan,

Thanks for your feedback!

I've made adjustments based on your suggestions:
- **feat(errors): implement SARIF support as a MessageStyle 
https://github.com/dlang/dmd/pull/17016)

https://github.com/dlang/dmd/pull/17020)

These should align the command-line interface and output format 
with GCC and Clang, and ensure external tools can work seamlessly 
with DMD+SARIF.

Cheers,
Royal
Oct 22