fixed problemmatcher regexp

[guntern]

Some gcc versions do not print the column number of an error or warning. These will not be catched by the problem matcher. I tweaked the regular expression so that it also matches, when the column number is not present.
The change in the first group is necessary so that it does not expand because of the greedyness to include the line number. It could also be replaced by (.*?) to have lazy matching.

[sean-mcmanus]

How do we repro gcc outputting warnings/errors without a column?

[bobbrow]

You can write a task that does an 'echo' of an error line without the column and set the problem matcher to this version of gcc.

[sean-mcmanus]

You can write a task that does an 'echo' of an error line without the column and set the problem matcher to this version of gcc.

Yeah, if required for testing, but I wanted to know how to repro the issue with gcc, i.e. what does "some gcc versions" mean?

[guntern]

Sorry, for not mentioning the version.
The problem ocurred for us with a gcc 3.4.6. This version produces output with and without column numbers. Two examples below:

File.c:416:10: warning: #warning test
File.c:423: error: [12874] conflicting types for 'test'

I just exchanged the file names.

I tested the regular expression on regex101.com until it worked as intended.

According to https://stackoverflow.com/questions/54541148/how-to-define-a-problem-matcher-for-cppcheck-task-in-vscode you can also reproduce the error when using cppcheck 1.82.

[sean-mcmanus]

[^:]* doesn't seems to work with Windows style file paths (e.g. c:\Users). This might be fixable via adding something like (.:)? before that. I think the original .* was not 100% guaranteed to work.

[guntern]

Indeed, Windows style full paths did not work. So I extended my test cases to

File.c:416:10: warning: #warning test
File.c:423: error: [12874] conflicting types for 'test'
C:\Path\to\file.c:423: error: [12874] conflicting types for 'test'
C:\Path\to\file.c:416:10: error: [12874] conflicting types for 'test'

I now changed the first group to (.*?), which is now doing lazy matching. This also prevents the line number to become part of the filename, if the column number is missing.

[sean-mcmanus]

C:\Path\to\file.c:416:10 error: [12874] conflicting types for 'test'

Is this test case valid (no colon after the comma) or should the regex (\d*):? be replaced with something like (?:(\d+):)??

[guntern]

C:\Path\to\file.c:416:10 error: [12874] conflicting types for 'test'

Is this test case valid (no colon after the comma) or should the regex (\d*):? be replaced with something like (?:(\d+):)??

Indeed, I missed the colon there. I updated the test cases in my comment accordingly.

However, my proposed version works with and without this colon. (?:(\d+):)? just matches when the colon is present. So it comes down to how strict the matching should be. Well there is one difference according to regex101: (\d*):? produces an empty group if it does not match. (?:(\d+):)? omits the group at all. I was afraid that this missing group would then cause errors in VS Code. But apparently it can handle this very well. So both versions work in VS Code and list the problems correctly.

So in my point of view both versions work, although slightly differently. It is up to you, which version you prefer.

[michelleangela]

(\d*):? that creates an empty group if the column number is not there seems to be a better choice to avoid potential errors in VS Code.

[sean-mcmanus]

(\d*):? that creates an empty group if the column number is not there seems to be a better choice to avoid potential errors in VS Code.

Yeah, that seems fine to me.

[sean-mcmanus]

Your fix is available with https://github.com/microsoft/vscode-cpptools/releases/tag/1.2.0-insiders2