Overlapping patterns in toplevel configure.in

Overlapping patterns in toplevel configure.in

[Andreas Schwab]

The toplevel congfigure.in script contains a case pattern *-*-linux* that
overrules the pattern mips*-*-linux*.  The first pattern only exists in
the gcc repository, not in the src repository.  There are also other linux
cases that are not a superset of the generic *-*-linux* case.  What is the
best way to resolve that?  Perhaps the generic linux pattern should be
moved to a separate case statement.  But there is also a catch-all
pattern at the end that disables libgcj for all not explicitly mentioned
targets unless --enable-libgcj is given.  In other words, it's a big
mess. :-(

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE GmbH, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Overlapping patterns in toplevel configure.in

[Thiemo Seufer]

Andreas Schwab wrote:
> The toplevel congfigure.in script contains a case pattern *-*-linux* that
> overrules the pattern mips*-*-linux*.  The first pattern only exists in
> the gcc repository, not in the src repository.  There are also other linux
> cases that are not a superset of the generic *-*-linux* case.  What is the
> best way to resolve that?

What about doing the minimum needed to fix it? The patch below moves
it downwards and updates the mips case.

> Perhaps the generic linux pattern should be
> moved to a separate case statement.  But there is also a catch-all
> pattern at the end that disables libgcj for all not explicitly mentioned
> targets unless --enable-libgcj is given.  In other words, it's a big
> mess. :-(

The check about rx is always the same and could be moved in a seperate
statement.


Thiemo


--- configure.in	Thu Mar 28 00:32:31 2002
+++ configure.in.mipsfix	Tue Apr  9 11:14:24 2002
@@ -796,11 +796,6 @@ case "${target}" in
          target_configdirs="${target_configdirs} target-libstub target-cygmon"
     fi
     ;;
-  *-*-linux*)
-    noconfigdirs="$noconfigdirs target-newlib target-libgloss"
-    # linux has rx in libc
-    skipdirs="$skipdirs target-librx"
-    ;;
   i[3456]86-*-mingw32*)
     target_configdirs="$target_configdirs target-mingw"
     noconfigdirs="$noconfigdirs expect target-libgloss ${libgcj}"
@@ -951,7 +946,9 @@ case "${target}" in
     target_configdirs="${target_configdirs} target-bsp target-libstub target-cygmon"
    ;;
   mips*-*-linux*)
-    noconfigdirs="$noconfigdirs target-libffi"
+    noconfigdirs="$noconfigdirs target-libffi target-newlib target-libgloss"
+    # linux has rx in libc
+    skipdirs="$skipdirs target-librx"
     ;;
   mips*-*-*)
     noconfigdirs="$noconfigdirs gprof ${libgcj}"
@@ -1023,6 +1020,11 @@ case "${target}" in
     ;;
   vax-*-*)
     noconfigdirs="$noconfigdirs target-newlib target-libgloss ${libgcj}"
+    ;;
+  *-*-linux*)
+    noconfigdirs="$noconfigdirs target-newlib target-libgloss"
+    # linux has rx in libc
+    skipdirs="$skipdirs target-librx"
     ;;
   *-*-lynxos*)
     noconfigdirs="$noconfigdirs target-newlib target-libgloss ${libgcj}"

Re: Overlapping patterns in toplevel configure.in

[Andreas Schwab]

Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> writes:

|> Andreas Schwab wrote:
|> > Perhaps the generic linux pattern should be
|> > moved to a separate case statement.  But there is also a catch-all
|> > pattern at the end that disables libgcj for all not explicitly mentioned
|> > targets unless --enable-libgcj is given.  In other words, it's a big
|> > mess. :-(
|> 
|> The check about rx is always the same and could be moved in a seperate
|> statement.

Actually, the statement is not true any more, unless we regard rx as
equivalent to GNU regex.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE GmbH, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Overlapping patterns in toplevel configure.in

[DJ Delorie]

> The toplevel congfigure.in script contains a case pattern *-*-linux* that
> overrules the pattern mips*-*-linux*.  The first pattern only exists in
> the gcc repository, not in the src repository.  There are also other linux
> cases that are not a superset of the generic *-*-linux* case.  What is the
> best way to resolve that?  Perhaps the generic linux pattern should be
> moved to a separate case statement.  But there is also a catch-all
> pattern at the end that disables libgcj for all not explicitly mentioned
> targets unless --enable-libgcj is given.  In other words, it's a big
> mess. :-(

I'm currently working on syncing the toplevel files between gcc and
src, so at least please wait until I finish that, or at least make
sure both sides get updated.  Otherwise, simply rearranging the
patterns so that the more specific ones come first makes sense;
perhaps put all the chip-independent patterns at the end?