Hi Paul, On Sat, 14 May 2011, Paul Eggert wrote:
> On 05/14/11 19:34, Joel E. Denny wrote: > > this should give distros a way out if they don't want > > to upgrade m4 and don't mind risking the bugs. > > We shouldn't recommend this, as the bugs can be > subtle and hard to debug. I ran into them when > configuring a test version of Emacs, and it wasn't fun. Good to know. Thanks. My fear is that some people will choose not to upgrade Bison to 2.5 because they will assume that, as implied by configure, 2.5 is somehow less able to cope with a buggy m4 than Bison 2.4.3 and earlier. I have not yet seen any evidence to support that assumption. Instead, those older Bison versions accept the buggy m4 versions only because the bugs hadn't been discovered (or the configure checks hadn't been written) by the time of those Bison releases. In conclusion, when Bison 2.5's configure rejects your m4, it likely is worthwhile to upgrade m4 regardless of whether you ultimately upgrade Bison to 2.5. That is, keeping your old Bison and your old buggy m4 is likely no better than forcing Bison 2.5's configure to accept your old buggy m4.
