reserved identifier violation

[elfring]

I suggest to try the search pattern "\b_+[A-Z]+" on the source files. You will find a couple of places where names are used that begin with an underscore and an uppercase letter.

Examples:

• _BOINC_API_
  
• _WIZ_ACCOUNTINFOPAGE_H_
  
• _BOINC_PROCESS

This does not fit to the expected ​naming conventions of the ​C language standard.

[Nicolas]

This report is about many different parts of the code, not API only.

[dandelot]

C++, sec 17.4.3.1.3: Each name having two consecutive underscores (2.11) is reserved to the implementation for use as a name with both extern "C" and extern "C++" linkage.

So various names like __cuInit defined-and-used in client/coproc_detect.cpp are not standard-conforming. AFAICT the identifiers used in coproc_detect.cpp are chosen by boinc, not cuda (my apologies for this note if those specific identifiers with leading underbars are somehow required by the cuda library, as in that case all is well).

Those seem to the only double-underscore names defined in boinc.

[elfring]

Replying to dandelot:

How do you think about a ​corresponding clean-up for your ​include guards?

[dandelot]

The include-guards rules (using guards that follow the C/C++ rules) suggested in the link make sense to me.

It's difficult to ensure that include-guards in boinc work without conflicting with other non-boinc identifiers, but adding to the referenced suggestions by preceeding (in boinc source) the guards with BOINC_ would be effective for most (assuming headers guards always had a trailing _H to avoid confusing them with capitalized class names like BOINC_PROCESS).

For example, in ViewNotices.h

ViewNotices.h:#ifndef _VIEWNOTIFICATIONS_H_

could be #ifndef BOINC_VIEWNOTIFICATIONS_H_ (of course change all instances of _VIEWNOTIFICATIONS_H_ so it all works.)

Having the header file named ViewNotices.h while the guard is renamed BOINC_VIEWNOTIFICATIONS_H_ (inconsistent spelling, notices vs notifications) is slightly jarring but otherwise harmless.

I hope these comments are responsive to elfring's question...