On Thu, Mar 11, 2010 at 07:18:45PM -0500, Charles Wilson wrote: >AFAIK, the new message: field in setup.hint is not intended to be a >required field. However, if it is missing from the setup.hint, genini >complains. > >Fine, I can work around this by (always...) using --okmissing=message. > >However, if I do that, then I get a setup.ini like the following: > >... >@ test-data >sdesc: "test package" >ldesc: "test package" >category: Utils >message: "" >version: 0.1-2 >install: release/test-data/test-data-0.1-2.tar.xz 116920 >e69dda662fde6b02644d4562c1ed0541 >source: release/test-data/test-data-0.1-2-src.tar.xz 108 >7f06fdbf4fd1eff94c1c85068c155c1c >... > >And when I run setup.exe, I get a popup with the following message: > >C:\cygwin-1.7\usr\src\devel\setup\test/setup.ini line 10: syntax error, >unexpected NL, expecting STRING >C:\cygwin-1.7\usr\src\devel\setup\test/setup.ini line 10: unrecognized >line 11 (do you have the latest setup?) > > >It appears there are three separate problems here. > >1) It'd be nice if genini didn't assume that all of the keywords it >knows about are required. E.g. pre-seed okmissing with (list of >optional keywords)? > >2) Failing that, since message is optional, then if the data structure >for a particular package has no entry for 'message', then...genini >should not print out an useless 'message: ""' line. > >3) Finally, setup.exe's behavior when seeing this entry...is this a bug? >If you have > message: "" >or even > message: >should setup.exe treat that like an empty requires:, and just ignore it? >Or do we assume that if a message: entry is present then there WILL be a >message popup...and it better have a non-empty string value?
genini shouldn't assume that message is required and setup.exe shouldn't have to care about blank messages. Btw, the syntax is (or should be): message keyword "something" That's why you're seeing a syntax error. cgf
