Re: glibc 2.0.95

[Franz Sirl <Franz.Sirl-kernel@lauterbach.com>]

At 17:02 30.07.98 , John Kennedy wrote:
>07/30/98 @ 08:02:31 AM (Thursday)
>
>[Thomas Quinot]
>>  It looks like seteuid is still broken. It may change the saved uid,
>>  which is not consistent with the 4.4BSD semantics.  This breaks at
>>  least Sendmail 8.9.*.
>
>  That is an old problem.  If you look in sendmail's "test/Results",
>you can see seteuid() problems going back to linux-1.2.13 (11/2/95).
>I actually tripped over that fairly recently since it mostly effects
>local->local email and submitted this patch to the sendmail folks:
>
>  The "POSIX" comment has to do with the way sendmail guesses default
>capabilities and why it favors seteuid() over setreuid() by default.
>
>  Should be fixed > sendmail-8.9.1.  (:

sendmail always needs fixing ;-)


>  As a side note, has anyone else tried to recompile egcs-1.0.3a under
>glibc-2.0.95?  In bootstrapping my homebrew distribution I got hung up
>in there with errors in egcs's libio.h header.  If I don't hear from
>anyone I'm going to try a egcs snapshot and then spot-compile a few
>more programs to see if it is egcs or a more general problem.  I just
>rebuilt from scratch the day before against linux-2.1.112, so moving
>from -2.0.94 to -2.0.95 should have been the only new variable.
>I get a lot of these:
>
>	/h/src/jk1/egcs-1.0.3a/BUILD/../libio/libio.h:270:
>		invalid exception specifications
>	/h/src/jk1/egcs-1.0.3a/BUILD/../libio/libioP.h:453:
>		invalid exception specifications
>
>  Where the line #s in libio.h & libioP.h are:
>	libio.h:	270 271 272 273
>	libioP.h:	94 100 107 119 126 133 143 156 164 170
>			176 194 201 209 219 226 233 238 453
>  
>  Those happen a fair way into the compile (libraries/libiberty).

These happen if a typedef uses __P instead of __PMT. Go and edit (or copy
from glibc?) egcs/libio/libioP.h and egcs/libio/strfile.h accordingly. Hope
this works for 1.0.3, it did work for 1.1.

Franz.