gogensig:panic when redefined

[luoliwoshang]

question 1

Tag names: all identifiers declared as names of structs, unions and enumerated types. Note that all three kinds of tags share one name space.

The following is valid because enum algorithm and typedef algorithm belong to different scopes:

typedef enum algorithm {
    UNKNOWN = 0,
    NULL = 1,
} algorithm_t;

typedef enum {
    UNKNOWN2 = 0,
    NULL2 = 1,
} algorithm_t2;

typedef algorithm_t algorithm;

If, in this case, the typedef algorithm is allowed to be mapped to another name, such as algorithm__1, it would be impossible to describe this mapping in the mapping file without causing ambiguity.

llcppg.pub

algorithm algorithm__1

The cname would correspond to two nodes: the enum node and the typedef node. Therefore, for this situation, identifying and skipping such cases would be a more appropriate choice.

typedef enum algorithm {
    UNKNOWN = 0,
    NULL = 1,
} algorithm;

In the main branch of llcppg, llcppsigfetch already supports filtering typedef self-references of this form, but now it needs to support multi-level self-references.

• llcppsigfetch:skip typedef multi-level self-references #286

question 2

The following is valid but currently causes a panic:

struct isl_arg_choice {
    const char *name;
    unsigned value;
};
enum isl_arg_type {
    // ....
    isl_arg_choice
    // ....
};

The reason is that struct isl_arg_choice and the enum item isl_arg_choice should exist in different scopes, but they are currently stored in the same map[cname]pubname without generating a new name with __x based on the node type.

• gogensig/refactor:collect name map by node #289

The isl_arg_choice as an enum item is still checked for whether its cname is in the current scope, used to submit a cname -> pubnamed type. However, for enum values, they exist as constant values and are not referenced as types in the code (e.g., syntax like isl_arg_choice x is invalid). Therefore, during the conversion process, we do not need to check the origin name of macros and enum items, but only verify their converted pubname.

• gogensig:lookup origin name for type decl #290

---

• Reduce unnecessary error interception.
• Instead of panicking directly within the package process, return the error and let the convert function handle it.

[codecov]

Codecov Report

Attention: Patch coverage is 92.50000% with 6 lines in your changes missing coverage. Please review.

Project coverage is 79.99%. Comparing base (d01d64b) to head (cb36331).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
cmd/gogensig/convert/convert.go	72.72%	2 Missing and 1 partial ⚠️
cmd/gogensig/convert/package.go	93.61%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #285      +/-   ##
==========================================
- Coverage   80.31%   79.99%   -0.33%     
==========================================
  Files          28       22       -6     
  Lines        2957     2919      -38     
==========================================
- Hits         2375     2335      -40     
+ Misses        567      564       -3     
- Partials       15       20       +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[qiniu-x]

[Git-flow] Hi @luoliwoshang, There are some suggestions for your information:

---

Rebase suggestions

• Following commits seems generated via git merge

  Merge branch 'gogensig/namemapper' into gogensig/panicredefine

  Merge remote-tracking branch 'upstream/main' into gogensig/panicredefine

Which seems insignificant, recommend to use git rebase command to reorganize your PR.

For other git-flow instructions, recommend refer to these examples.

If you have any questions about this comment, feel free to raise an issue here:

• https://github.com/qiniu/reviewbot