On Thu, 08 Jan 2004 10:27:17 +0100, Alexandre Duret-Lutz wrote: > >>> "Eric" == Eric Sunshine <sunshine@xxxxxxxxxxxxxx> > Eric> Nevertheless, DOS will silently truncate those to > Eric> "config.sta" and "configur", both of which are still > Eric> valid filenames. However, the second period in > Eric> "config.status.lineno" is simply incompatible with DOS > Eric> naming convention which allows only a single period. > Is this file created on 8+3 platforms? For instance I believe > DJGPP uses bash, which supports $LINENO. I suspect there might > be no issue. I do not know specifically whether config.status.lineno is created on any 8.3 platforms. I was just asking out of general curiosity since it struck me as inconsistent that it unconditionally uses this non-8.3-conformant name whereas the rest of Autoconf tries to cater to 8.3 platforms by providing alternate mechanisms for dealing with those platforms (even if those mechanisms are not used by default). You are probably right about DJGPP using Bash, so it may not be an issue in actual practice. (Although this does not mean that we shouldn't fix it.) > >>> "Daniel" == Daniel Reed <n@xxxxxx> > Daniel> There is already some precedent for ignoring such > Daniel> restrictions: Autoconf's autoheader will happily create > Daniel> src/config.h.in given an AC_CONFIG_HEADER of > Daniel> src/config.h and no explicit infile. > We should just ensure that users _can_ write a package that > build on such platform. Using config.hin instead of config.h.in > is documented, many people do it. Exactly. Even though Autoconf uses non-8.3 names by default, it still tries to cater to 8.3 platforms by providing alternate mechanisms for dealing those platforms; and these alternate mechanisms can be utilized/enabled as needed. -- ES
