Include version_enforcer.h in all headers in CGAL

[janetournois]

@CGAL/geometryfactory what would you think about including the version_enforcer.h file to every header in CGAL, and adding the corresponding file to every release?
It would sometimes help find inconsistencies between the versions that are used, in particular in specific projects development.

Then, in projects outside the repo, use a local version_enforcer.h, to make sure the user is using the right version of CGAL.

[lrineau]

Why not include it in <CGAL/config.h>?

[MaelRL]

Can you please illustrate the first use case? The file uses a hard-coded version number that is checked against the CGAL version macros; if you put the file within CGAL, then the macro must have the version number, and then what is its purpose?

[janetournois]

@MaelRL In the "normal" setup it is useless.
When you start using copied/forked header files, it would be nice to be able to check at compile time whether the forked files (still) "belong to" the same version of CGAL as the version of CGAL read in CGAL_DIR (without introducing an extra-fork by including version_enforcer.h)

@lrineau it sounds good... I'm wondering if it covers all our fork-cases

[sloriot]

To do that you need to have a version macro in each file, including a header does not help.

[MaelRL]

If you're forking specific files, you'd have to tag them with something that lives entirely within the forked file (or in your patch repository) since you can't "trust" the rest of the CGAL headers.

You could define a version macro in your forked files and have the verison_enforcer.h in CGAL (config.h or somewhere else), but that's equivalent to putting a version_enforcer.h in the patch repo.

[sloriot]

Indeed you need a macro per forked file.

[janetournois]

My idea is

• add a straightforward version_enforcer.h in the CGAL repo, and include it everywhere (maybe through config.h)
• add a specialized version_enforcer.h in my fork, that is included BEFORE the one of CGAL, and pre-defines the acceptable CGAL version

[janetournois]

The aim is to avoid adding #include <version_enforcer.h> in each forked file (because the include is already there in the original CGAL file)

[lrineau]

Jane's proposal

Let me clear what I have understood. The goal of Jane is to enforce the same version in two files:

• <CGAL/version.h>, and
• <CGAL/version_enforcer.h>.

By default, without any user-specific modified headers, both would come from a distribution of CGAL, and would of course agree.

With user-specific modified headers, a specific CGAL/version_enforcer.h would be shipped with the modified headers, and it will ensure that the version of the CGAL distribution is the one expected by the modified headers.

@janetournois: Is that a correct rephrasing of your idea, Jane ?

Mael's proposal

Another proposal from Mael would be to encode the CGAL version in all CGAL headers, and make sure they can only be used in the right CGAL distribution.

[MaelRL]

No, this isn't what I'm saying. I was just trying to understand Jane's proposal (specifically the first part, where there is no forked files and where the file is useless then).

I don't have any strong objection to this mechanism to automatically include the version enforcer, but the version_enforcer.h (whichever its implementation) should become obsolete rather soon anyway since we will not have forked files living outside CGAL releases.

[janetournois]

that's right...

[lrineau]

The cost of adding a trivial header version_enforcer.h and

#include CGAL/version_enforcer.h>

in <CGAL/config.h> is near zero, anyway. I propose we do that.

We can even add it in CGAL-5.6, because it is really harmless.

[lrineau]

@janetournois Please create the pull-request. We will merge it.

[janetournois]

@lrineau yes, exactly.

It seems that Mael's proposal is another solution to the same problem.

[lrineau]

The cost of having the CGAL version in all CGAL headers is high: for each new release, that would mean modifying more than 3400 headers. We can do it, but if we are sure of the benefit.

[janetournois]

Done in PR #7528