> the problem for the case that fails is that the iffe translation of
> features/locale => FEATURE/locale
> does not produce
> #define _typ_wctrans_t 1 /* wctrans_t is a type */o

I can confirm this. 
$ grep wctrans .build-20110104-1.fc15.log 
iffe: test: is wctrans a library function ... yes
iffe: test: is towctrans a library function ... yes
iffe: test: is wctrans_t a type or typedef ... no
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/src/cmd/ksh93/sh/init.c:2087:4: 
warning: passing argument 2 of 'towctrans' makes pointer from integer without a 
cast
/usr/include/wctype.h:222:15: note: expected 'wctrans_t' but argument is of 
type 'sh_wctrans_t'

> the mystery is that the iffe.out sent by you and a previous one
> reported by Olga have this define, so there is a diff between the
> manual
> iffe and the one run by the original build
> 
> can you run the iffe command with the shell used in the original build
> 
> SH=path-to-the-original-sh
> SHELL=$SH $SH
> iffe -v -d2 - typ wctrans_t wctype.h > iffe.out 2>&1
> 
> if that produces the #define 

it does

> then snarf the iffe*features/locale line
> from the original build make.out

do you mean this line?

iffe -v -c 'cc  -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions 
-fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic -O0 -ggdb3    ' 
ref -L. -L/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib 
-I/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/include/ast 
-I/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/include 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libdll.a -ldl 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libast.a 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libcoshell.a 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libast.a 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libcmd.a 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libast.a 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libast.a -lm : 
run /mnt/home2/mhlavink/gitf/ksh/ksh-20110104/src/cmd/ksh93/features/locale

in make.out is:

ffe: test: is sys/types.h a header ... yes
iffe: test: is 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libdll.a a 
library ... yes
iffe: test: is -ldl a library ... yes
iffe: test: is 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libast.a a 
library ... yes
iffe: test: is 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libcoshell.a a 
library ... yes
iffe: test: is 
/mnt/home2/mhlavink/gitf/ksh/ksh-20110104/arch/linux.i386-64/lib/libcmd.a a 
library ... yes
iffe: test: is -lm a library ... yes
iffe: test: is locale.h a header ... yes
iffe: test: is wchar.h a header ... yes
iffe: test: is wctype.h a header ... no
iffe: test: is locale a library function ... no
iffe: test: is localeconv a library function ... yes
iffe: test: is wctype a library function ... yes
iffe: test: is iswctype a library function ... yes
iffe: test: is iswblank a library function ... yes
iffe: test: is wctrans a library function ... yes
iffe: test: is towctrans a library function ... yes
iffe: test: is time.h a header ... yes
iffe: test: is sys/time.h a header ... yes
iffe: test: is sys/times.h a header ... yes
iffe: test: is stddef.h a header ... yes
iffe: test: is stdlib.h a header ... yes
iffe: test: is wctrans_t a type or typedef ... no
iffe: test: cat{ ... }end ... yes

and from manual execution, the output is the same

I don't know why it reports "is wctype.h a header ... no" because wctype.h 
exists in expected directory: /usr/include/wctype.h

> and run that -- maybe there is
> an unexpected interaction (there shouldn't be) when running the
> full iffe script vs the single test above
> 
> On Thu, 6 Jan 2011 15:21:15 +0100 Michal Hlavinka wrote:
> > > Michal, that is a bug in one of the iffe probes. The bug will only
> > > happen for platforms with sizeof(int) !=3D sizeof(long), i.e.
> > > 64bit
> > > Linux machines. Solaris is OK.
> > >=20
> > > We've observed this on SuSE 11 but more data, e.g. from Fedora may
> > > help. Please run this and send iffe.out to this list:
> > >        bin/package use
> > >        cd ksh93
> > >        iffe -v -d2 - typ wctrans_t wctype.h > iffe.out 2>&1
> 
> > attached=20
_______________________________________________
ast-developers mailing list
[email protected]
https://mailman.research.att.com/mailman/listinfo/ast-developers

Reply via email to