Public SourceMage Discussion List

["M.L." <mlubrano AT sourcemage.org>]

On Mon, 16 Dec 2002 20:35:15 +0800, Jimmy Yen(???) wrote
> On Fri, Dec 13, 2002 at 11:09:14AM -0800, Eric Sandall wrote:
> >/ Well, apparently slang has the same problem as ncurses, with regards to
> />/ gcc installing their own version.
> />/ 
> />/ I recall someone saying (or gainsaying, don't recall) that 
> recompiling gcc />/ after ncurses is updated would fix the ncurses.h 
> file which gcc installs, />/ so that we'd no longer have to remove 
> it after an ncurses update (it is a />/ Bad Thing (TM) to remove 
> files from a spells installation, makes them />/ "corrupted" for 
> 'cast --fix'). />
> >/ So, if recompiling gcc after ncurses (and slang) fixes those, we should
> />/ add triggers to gcc so that it recompiles after these are 
> updated (and any />/ others that people find).  I haven't tried 
> recompiling gcc to see if this />/ works (recompiling now, but it's 
> a PII-300, so it'll take a while.  ;)), />/ but if anyone else 
> knows/can try faster, please do. />/ />/ Here is the slang info: />/ 
> http://bugs.sourcemage.org/show_bug.cgi?id=1957 />/ />/ Thoughts? 
> />/ />/ -One of Four />/ a.k.a. sandalle/
> 
> It seems that the cause of all the problems is that gcc examines and
> "fixes" all the header files in /usr/include and its subfolders when 
> it compiles itself, and keeps modified ones in a seperate folder. 
> When a spell installs a new header which is not compatible with the 
> old one, programs surely won't compile.
> 
> The problem is that gcc does the "fixing" job only once, when itself 
> is compiled. The only way to get newly installed headers "fixed" 
> (and thus updating gcc's modified headers) is to recast gcc. But 
> since we don't know which header files gcc will deem "flawed", we 
> will have to recast gcc constantly to make it "fix" newly installed 
> header files.
> 
> It came to me that it may be possible to check the file names of the
> header files each existing spell updates, and recast gcc if there's a
> match with its modified header files. But then we still need to cast
> gcc very often (that's apparently not a good thing,) and if a new spell
> is installed, the headers they install will only get fixed next time
> gcc compiles. That will run the fixing mechanism of gcc worthless.
> 
> IMHO, hiding the gcc-modified headers when casting certain spells is
> simply wrong, because we don't either use the gcc-modified headers, nor
> updating them. I think we should either recast gcc constantly (which
> seems really impractical,) or don't let gcc store the fixed headers 
> at all. Removing the gcc-modified headers directly when casting gcc can
> make our life a lot easier, if it's OK to do so.
> 
> (Uh, sorry for my English. Hope you can get my point. Also, because I'm
> not far from a newbie, please forgive me if I'm seriously wrong.;)
> 
> haoto
> 
> _______________________________________________
> SM-Discuss mailing list
> SM-Discuss AT lists.ibiblio.org
> http://lists.ibiblio.org/mailman/listinfo/sm-discuss

What if you make symlinks from unmodified headers over gcc's fixed headers ?


//
//   M.L.
//   Linux User #36502 ; ICQ #23832402
//