On Fri, 18 Nov 2022 11:27:14 GMT, Magnus Ihse Bursie <i...@openjdk.org> wrote:
>> Julian Waters has updated the pull request incrementally with one additional >> commit since the last revision: >> >> Uh oh > > Aaaahhh, **that** bug. I've run into that before. I did attempt to fix it but > after wasting too many hours I gave up. :-( It seems to be a limitation in m4 > that I cannot understand how to get around. At some point, the string > literal, even though quoted inside the `[...]`, is expanded and parsed as m4 > macro expansion arguments. In theory, I should have been able to add an > additional layer of quoting and then un-quote it once it was past the > problematic point, but that did not work. > > I solved the problem by not solving it, and instead rephrased the message to > not need the comma. (Imho, this limitation actually improved the quality of > the descriptions, so it was not bad per se). > > But I agree that it is annoying to have such a limitation in > `UTIL_DEFUN_NAMED`. If you want to have a go at trying to solve it, please > do! I'll fully admit my shortcomings and state that trying to solve this > passes my knowledge and ability to manipulate m4. > > Or, maybe, you could add some documentation to `UTIL_DEFUN_NAMED` and > `UTIL_ARG_WITH`, saying that comma is not allowed in the values. @magicus I need some of your input on this: Typically when using AC_ARG_WITH one would surround the if block with an extra quote, for instance if we had `if ! [[ $JDK_RC_NAME =~ ^[[:print:]]*$ ]]` the code in m4 would have an extra set of brackets to quote it so m4 doesn't absolutely wreck the final script that we want to emit, such as `if [ ! [[ $JDK_RC_NAME =~ ^[[:print:]]*$ ]] ]` I'm about halfway through fixing the whole issue with commas, and I'm fairly certain I could lump the problem with quoting expansion in there too. My question though is: When it comes to parameters that take "predicate" code blocks as arguments (Such as CHECK_VALUE), should I have UTIL_ARG_NAMED handle the annoying issue of quoting entirely so the caller can freely pass `if ! [[ $JDK_RC_NAME =~ ^[[:print:]]*$ ]]` without having to worry at all, or should I make users still have to keep passing that extra level of quotation into things like IF_GIVEN to match what is already being done? ------------- PR: https://git.openjdk.org/jdk/pull/11020
