Nothing is working, where did I go wrong?

Sorry to hear that. Please try comparing how your documentation is set up with the companion website.

If things look similar enough, or something isn’t clear, raise an issue on GitHub. I’ll do my best to support what I can, and if similar questions come up then I can add them to this FAQ.

Make sure you set the verboseBuild to True, there may be some useful information being printed there.

Can I use the formidable Intersphinx extension?

Heck yes! This has almost nothing to do with Exhale, and everything to do with Sphinx. I’ve prepared a crash-course on how to get it up and running on the companion site’s Using Intersphinx page.

Why does it build locally, but not on Read the Docs?

Most likely Exhale is failing to build if you are getting this. Check the build logs on RTD, and also make sure that you are using the correct version of python you desire by configuring the read the docs yaml.

Note that you may specify a version of python, as well as a path to a requirements.txt to have your project’s documentation dependencies installed.

Why are there build warnings about generated files not being included?

This means that you have changed your API in a way that the files being generated have changed. For example, you changed the template parameters on something, or you deleted a class. So now when you run make html, the original generated files are still floating around down there even though the current execution of Exhale will not use them.

Solution: see the Optional: Create a Proper Clean Target section of the Quickstart guide.

Unicode support?

Every action Exhale performs with respect to strings is done using unicode strings. Or at least I tried my very best to make sure unicode support works.

  1. At the top of every file:

    from __future__ import unicode_literals

    This is why all of the documentation on this site for strings has a leading u.

  2. Every file written to:

    with codecs.open(file_name, "w", "utf-8") as f:
  3. When using Python 3, bytes conversion is done as:

    doxygen_input = bytes(doxygen_input, "utf-8")

The last one may potentially cause problems, but in local testing it seems to be OK. E.g., if you only specify ASCII in your conf.py, everything should work out in the end.


Sphinx / reStructuredText supports utf-8 no problem. The only potential concern is communicating with Doxygen on stdin like this, but it’s worked without issue for me so I kept it. Please speak up if you are experiencing encoding / decoding specific issues in Exhale!

Why does my documentation take so long to build?

This is a byproduct of what is actually being done by Exhale. If you look at the build output of Exhale when you execute make html, parsing and generating the documents takes on the order of seconds.

What takes long is Sphinx, and the time it takes is directly proportional to the size of the API being documented. The larger the API, the more individual reStructuredText documents there are being created. Meaning there are more documents that Sphinx has to read and write.


The sphinx-bootstrap-theme is noticeably slower than others. I have suspicions as to why, but have not actually investigated potential fixes.

Metaprogramming and template (specialization)?

Yes and no. Partial and full template specialization are supported, but not elegantly.

  1. Currently there are no links from partial and full specializations back to their original (unspecialized) type. This may change in a future release. Finding the unspecialized type is complicated due to how things are presented by Doxygen.

  2. Template classes / structs were given the most attention. Functions may or may not work.

  3. All template classes, specialized or not, produce build warnings. These warnings come from Breathe. The documentation appears, but the layout is a little strange. For specializations in particular, they seem to produce an extra template <> in the output.

    For example, with a template <typename T, unsigned int N> DerivedClass,

    Partial Specialization template <unsigned int N> DerivedClass<int, N>

    Produces template <unsigned int N> template<> DerivedClass<int, N>.

    Full Specialization template <> DerivedClass<bool, 2>

    Produces template <> template<> DerivedClass<bool, 2>

  4. Where metaprogramming is concerned, it is more likely that Doxygen’s preprocessor needs to have everything PREDEFINED. YMMV.


If all else fails, you can force Doxygen to skip things. See the next FAQ entry.

How can I get Doxygen to skip code it cannot process?

It depends on what you need. If it’s something like a macro that isn’t expanding correctly, you can try pre-defining it. Otherwise, you skip it with preprocessor symbols that are only defined when the documentation is building. See the Doxygen PREDEFINED section.

I thought reStructuredText was supported, why does it look weird?

If you’re using complicated syntax (e.g., more than **bold** or listings), you will likely want to put that documentation in a raw reStructuredText verbatim block. This basically tells Doxygen to skip it, allowing for Breathe / Exhale to then process it directly.

See the Doxygen ALIASES section for how to do this.

The likely problem: Doxygen runs first. It supports Markdown, and it is probably transforming your documentation based off Markdown rules before Breathe / Sphinx / Exhale even gain access to it. Forcing a verbatim reStructuredText environment means that Doxygen simply passes the raw docstring unadulterated forward.

You aren’t including my enum’s, defines, etc. Why Not?

This happens because Doxygen is not including them. See the File Documentation is Necessary for More than just Files! section.

Why aren’t my Classes, Structs, Enums, or Unions associated with the right file?

I’m not entirely sure. Fortunately, you can specify the path explicitly for these. See the Associating Documentation with the Right File section. Personally, I tend to just default to always specifying the path manually.

My documentation is setup using groups, how can I use Exhale?

I do not support groups with Doxygen, as I assume if you have gone through the effort to group everything then you have a desire to manually control the output. Breathe already has an excellent doxygengroup directive, and you should use that.

How do I modify the pygments highlighter for a program listing?

By default Exhale will use utils.LANG_TO_LEX to choose the pygments syntax highlighter for .. code-block:: <lexer>. For most projects the defaults here should work as expected. If you need to change it, set lexerMapping in your conf.py.

Ok seriously, why is there SO MUCH documentation on a documentation system?

It’s your choice whether or not you read it. Back when I was getting started on all of this stuff it was overwhelming. I did my best to recall where I got confused, as well as highlight some of the tricks I’ve picked up over the years.

Basically, it’s because I’m hopeful that I can save at least one person from falling into some of the more confusing “traps” I’ve encountered.