Jim Meyering wrote: > Eric Blake wrote: >> On 01/17/2012 11:44 AM, Jim Meyering wrote: >> >>>>> * doc/INSTALL: Likewise. >>>>> * doc/INSTALL.ISO: Likewise. >>>>> * doc/INSTALL.UTF-8: Likewise. >>>> >>>> These files are generated through doc/Makefile. Also, on text files >>>> that consist of multiple paragraphs, I find it good (esthetic) to have >>>> a blank line at the end. It helps when you do "cat INSTALL" or >>>> "more INSTALL" (especially with the BSD 'more' program that terminates >>>> when it has printed the last line). >>> >>> I already exempted all of modules/* for you (I prefer to omit >>> the trailing blank line everywhere). Exempting three more is no problem. >>> Well, not much of a problem. Exempting them would make the regexp go >>> beyond column 80, so this splits it. Not pretty: >>> >>> diff --git a/cfg.mk b/cfg.mk >>> index bce58d2..2b4bcc8 100644 >>> --- a/cfg.mk >>> +++ b/cfg.mk >>> @@ -48,8 +48,9 @@ local-checks-to-skip = \ >>> sc_unmarked_diagnostics \ >>> sc_useless_cpp_parens >>> >>> +empty-at-EOF = >>> /fr/LC_MESSAGES/test-quotearg\.mo$$|^doc/INSTALL(\.(ISO|UTF-8))? >>> exclude_file_name_regexp--sc_prohibit_empty_lines_at_EOF = \ >>> - ^modules/|^lib/javaversion\.class$$|/fr/LC_MESSAGES/test-quotearg\.mo$$ >>> + ^modules/|^lib/javaversion\.class$$|$(empty-at-EOF) >> >> This part of your commit looks fishy: you touched up doc/INSTALL and >> friends, but the next time doc/install.texi is changed, they will be >> regenerated. So right now, the exemption for those files does nothing, >> but the generated files are also out of sync. >> >> I would be in favor of Paul's proposed patch that makes the generation >> omit the trailing blank line in the first place, but if we do that, then >> we _also_ need to remove doc/INSTALL from the empty-at-EOF exemption. > > I too would prefer to use Paul's patch and thus to remove the > doc/INSTALL* exemptions. > > Bruno, I know it's not your preference, but would that be ok with you.
Something else to consider: We are version-controlling these generated INSTALL* files. That is already on the edge of acceptability, but we're making exceptions here so that those files are easily accessible via the existing gitweb interface. IMHO, especially in *generated*-VC'd files, we can make an effort to have them conform.
