XTypes Branch: lint.pl: Additional Checks and Improvements

[iguessthislldo]

• Added the following checks and changed OpenDDS to pass these checks:

  • whitespace_before_newline_in_string: ex: "This is a string \n"

  • missing_include_guard: Checks IDL and header files for include guards matching one of two patterns. Both start with the file path, take dds/ off the front, turn anything that's not alphanumeric or an underscore to underscores, uppercase everything and prepend OPENDDS_. The difference between them is one converts from camel case to snake case while the other doesn't. For example, dds/DdsDcps.idl can be OPENDDS_DDSDCPS_IDL or OPENDDS_DDS_DCPS_IDL.

  • nonrelative_include_path: Checks to see if an include statement path contains the including file's directory in it. For example if foo/bar.h has #include <foo/util/stuff.h>, it should be #include "util/stuff.h".

• Added --try-fix to try to fix issues. So far this only works for the checks with strip_fix property set and the missing_include_guard check if there's an existing include guard.

• Way to pass arguments to ACE fuzz.pl.

• Add more info to --help message.

• 🌈 ANSI Escape Code Colors in messages 🌈 Can be controlled using command line arguments and environment variables.

• For checks using line_matched that support it, highlight what the offending characters of the line are in a separate error message.

• Rename not_exactly_one_eof_newline check to eof_newline_count. Made it accept a blank line at the end of text files. Also fixed an issue where it actually wasn't catching when there were no newlines at the end of the file. This means there must be 1 or 2 newline characters at the end of a file now.

• Checks are now listed and happen in consistent order now (alphabetically) and a proper error message is displayed when an invalid check is passed as an argument.

• Ability to check paths outside of OpenDDS, assuming the check isn't OpenDDS specific.

[jwillemsen]

Running this kind of generic cleanup on the xtypes branch could result in more merge conflicts on master, wouldn't it be better to run this on master and merge that in xtypes to prevent merge conflicts?

[iguessthislldo]

I did consider that, but like in #2114, I'm trying to avoid putting these kinds of changes in master since we're trying to do a release. Plus doing it here means I can make sure the xtypes branch is compliant, instead of waiting until mater is merged in.

[iguessthislldo]

I don't know what's going on in GCC6 yet, but the Java support has revealed an issue in tao_idl. tao_idl doesn't run the preprocessor on the IDL file itself, but on a temporary copy it creates with a cpp extension. I don't know how it is with msvc or other compilers, but it looks like it uses this to force gcc and clang to see the file as something to run the preprocessor on. Since it's in another location it breaks relative includes unless things happen to be set up correctly.

The way I see it there are at least three options:

• Change tao_idl to stop creating a copy of the IDL file and use something like gcc -x c instead. I would like to do that because it seems like the proper thing to do, but I'm not sure how compilers other than clang will work with this.
• Change tao_idl to add the directory of idl file to the include path. This is probably safer, however it will just simulate relative includes and won't work if trying to include staring with ...
• Workaround it. Basically I would revert the IDL files changes for this and exclude IDL files from the nonrelative_include_path check. To be safe I'd do the same thing in Simplify Includes DOCGroup/ACE_TAO#1344. I'd prefer to avoid this.

[iguessthislldo]

The Java issue is fixed with ACE/TAO #1357.