Comment author: @gasche

I suspect that this is an unintended change (I would expect all comments to be stored in this field). Unfortunately it's probably too late to fix this regression in a 4.02.x release.

Comment author: @lpw25

I cannot reproduce the problem. Could you give more details about what differences you are seeing?

Comment author: @trefis

The only difference I could notice between 4.02.1 and >=4.02.2 is that before 4.02.2 an ocamldoc comment would be stored with a leading star, whereas now you don't have that star anymore (since '(**' is now handled specially).

If you tool relied on that, maybe that explains the breakage?
In any case, I can confirm that all comments are still stored in the cmt_comments field.

Comment author: @oandrieu

Ooh right, indeed it's the missing leading star that broke my tool.

I had then a quick look at the patch, misread it and thought the comments were gone from cmt_comments.

sorry about that.

Comment author: @alainfrisch

I'm not sure about the resolution status. If the source code contains a comment (** ... *), shouldn't the leading star be reflected in the cmt_comments field?

Comment author: @lpw25

If the source code contains a comment (** ... *), shouldn't the leading star be reflected in the cmt_comments field?

I had thought not, but I guess there is no other way to distinguish special and regular comments in that field.

Comment author: @alainfrisch

My question was rather rhetorical. Since comments are delimited by (...), it seems pretty clear that the "payload" of (*FOO) is "*FOO". The fact that ocamldoc-like comments are recognized in a special way and passed through attribute is an addition, it should not break existing tools that rely on finding "*FOO" in the cmt files (esp. when the new -keep-docs is not used).

Comment author: @gasche

This was fixed by Leo in

#276