[regression] cdac1a6 broke successive alignments of a same range in some cases

[sim590]

Let's say you have this starting text:

int thisVariable = callingThisFunction(thisArg, thisOtherArgument, 0);
                   callingThisFunction(noThisArgumentNowHere, nowThisArg, 0);
                   callingThisFunction(thisArg, hereIcomeWithAnotherArgument, 0);

since cdac1a6, if you use GTabularize argument_list, then you get:

int thisVariable = callingThisFunction(thisArg,               thisOtherArgument,            0);
callingThisFunction                   (noThisArgumentNowHere, nowThisArg,                   0);
callingThisFunction                   (thisArg,               hereIcomeWithAnotherArgument, 0);

instead of the proper:

int thisVariable = callingThisFunction(thisArg,               thisOtherArgument,            0);
                   callingThisFunction(noThisArgumentNowHere, nowThisArg,                   0);
                   callingThisFunction(thisArg,               hereIcomeWithAnotherArgument, 0);

which you do get on the parent of cdac1a6.

I don't know what cdac1a6 does fix really, but I have not noticed any clear advantage since I updated to cdac1a6, but I got this really annoying regression.

Since there is no mention of an issue linked to cdac1a6 supporting it, I'm wondering if it is really useful. The explanation in the commit only states that the approach to indentation has changed, but what does it fix?

For now, I will config my Vimrc to use tabular at the version prior to cdac1a6.