I just disabled the module kind tag generation so these tags aren't generated right now. But this also means that for (sub)modules created this way: ``` module ModuleFoo = struct type foobar = ConstructorFoo | ConstructorBar of int * char list end ``` the tag `ModuleFoo` isn't created either and then we get `ModuleFoo.foobar` in the symbol tree because the parent node `ModuleFoo` doesn't exist. But I think it isn't such a big deal.

Oh, or do you mean how it works in universal-ctags?

Yes

There the unit tests work differently and preserve the name of the tested file so the output is always the same.

Ok, I thought there were other languages that required modules to be in files of the same name, but maybe ctags doesn't parse modules for those.

with our unit tests the source file name is a temporary name

Where does that happen? AFAICT `runner.sh` doesn't copy the source file, it just puts the tagfile in a temporary directory?

Or do you mean that by convention our test file names are chosen to reflect the test purpose (hence your choice of `simple.ml` for the first test file)? If so thats just a convention, and if it clashes with the language requirements it can be changed and your test file called `modulefoo.ml`.

Personally I would have disabled the tests not the parsing.

Note that the code is very c/c++ specific

To be fair the manual basically says that :stuck_out_tongue_winking_eye:

But if Geany is run with -P it doesn't use the cpp, but of course can only do one file at a time. Still that should do for ctags testing shoudn't it?

But if Geany is run with -P it doesn't use the cpp, but of course can only do one file at a time. Still that should do for ctags testing shoudn't it?

No, it can be all the files passed on the command-line (including glob patterns) - they would just be parsed one by one and the resulting tags combined. This is what I did in the just pushed commit. There's the small difference in the result of parsing anonymous tags described in the commit message but this isn't something we care about in global tags.

What part(s) does the no apply to?

(2) is correct ;-) (i.e. it is still possible to get tag file for multiple source file even with `-P`). To make things clearer, for

``` geany -g file1.h file2.h file3.h ```

Geany creates `tmp.h` containing ``` #include "file1.h" #include "file2.h" #include "file3.h" ``` and makes the preprocessor perform the includes and Geany then parses the resulting file (this is valid only for C/C++, for other languages Geany always behaves as if `-P` were used). The generated header file is syntactically valid, C/C++ doesn't care about file name so there's no problem with this method.

With ``` geany -P -g file1.h file2.h file3.h ``` Geany used to concatenate the files and parse the resulting file. After this patch, Geany parses each of the files separately, combines the resulting tags, sorts them and removes duplicates. The should be the same except for the minor difference for anonymous tags described in the previous post which isn't important for global tags.

I really don't see any reason for performing the concatenation - first, there's a need for the temporary file which may be a problem for languages like Ocaml, but second, it may even lead to a syntactically invalid file for some languages. For instance Go programs start with package specification followed by imports and these are not allowed to appear later in the source. So when you concatenate two files like ``` package main

import "fmt"

func main() { fmt.Println("Hello, 世界") } ``` the result is invalid. It then depends on how the ctags parser is constructed and if it is tolerant enough to allow that - it may, it may not but we are risking problems.

Oh sure, there are even more important things in the world than the tag manager :-).

There's both the Ocaml parser and the change not to concatenate parsed files in this PR. Should I split these into two separate PRs?

I created #3204 with the "non-concatenation" patch and will update this PR when/if it gets merged.

Since #3204 is already merged, I rebased this PR on top of master.

@techee pushed 1 commit.

c509b9000633acd9c9af93f15a84a3f4e6dd5ad3 Add Ocaml ctags parser

I'll look at the details later, but sure it looks like a worthy improvement for little expanse :+1:

Oops :) But yay, /me learnt a word

Merged #3163 into master.