[PATCH] configure: replace ${CXXLAGS} with ${CXXFLAGS_for_sh}

[PATCH] configure: replace ${CXXLAGS} with ${CXXFLAGS_for_sh}

[Tomi Ollila]

Variable CXXLAGS expands to nothing, CXXFLAGS something unusable
here; CXXFLAGS_for_sh expands to what we expect here.
---

I was palying with a patch set that enables 'set -eu' in usable way
when I noticed this...

 configure | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/configure b/configure
index 09e009b34115..0cbd64b1d26b 100755
--- a/configure
+++ b/configure
@@ -380,7 +380,7 @@ int main(int argc, char** argv) {
    Xapian::WritableDatabase db("test.db",Xapian::DB_CREATE_OR_OPEN);
 }
 EOF
-    ${CXX} ${CXXLAGS} ${xapian_cxxflags} _default_backend.cc -o _default_backend ${xapian_ldflags}
+    ${CXX} ${CXXFLAGS_for_sh} ${xapian_cxxflags} _default_backend.cc -o _default_backend ${xapian_ldflags}
     ./_default_backend
     if [ -f test.db/iamglass ]; then
 	default_xapian_backend=glass
-- 
2.6.4

[PATCH] configure: add set -u

[Tomi Ollila]

In case of any unset variable, make ./configure exit with nonzero value;
an attempt to expand an unset variable is a bug in the script
(usually a spelling mistake) and those should not pass through
unnoticed.
---

I was also doing set -e support but that has quite a few nontrivial
things in it. Set -u catches all lagging c++ compilers and so on ;).
Perhaps later.

This change requires id:1462296224-768-1-git-send-email-tomi.ollila@iki.fi
to be applied first, otherwise ./configure will not pass.

 configure | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/configure b/configure
index 6231d2b..3753e7f 100755
--- a/configure
+++ b/configure
@@ -1,5 +1,7 @@
 #! /bin/sh
 
+set -u
+
 # Test whether this shell is capable of parameter substring processing.
 ( option='a/b'; : ${option#*/} ) 2>/dev/null || {
     echo "
@@ -480,6 +482,7 @@ if pkg-config --exists valgrind; then
 else
     printf "No (but that's fine).\n"
     have_valgrind=0
+    valgrind_cflags=
 fi
 
 printf "Checking for bash-completion (>= 1.90)... "
@@ -490,11 +493,11 @@ else
     WITH_BASH=0
 fi
 
-if [ -z "${EMACSLISPDIR}" ]; then
+if [ -z "${EMACSLISPDIR-}" ]; then
     EMACSLISPDIR="\$(prefix)/share/emacs/site-lisp"
 fi
 
-if [ -z "${EMACSETCDIR}" ]; then
+if [ -z "${EMACSETCDIR-}" ]; then
     EMACSETCDIR="\$(prefix)/share/emacs/site-lisp"
 fi
 
-- 
2.8.2

Re: [PATCH] configure: add set -u

[David Bremner]

Tomi Ollila <tomi.ollila@iki.fi> writes:

> In case of any unset variable, make ./configure exit with nonzero value;
> an attempt to expand an unset variable is a bug in the script
> (usually a spelling mistake) and those should not pass through
> unnoticed.

pushed these two to master

d

Re: [PATCH] configure: add set -u

[David Bremner]

Tomi Ollila <tomi.ollila@iki.fi> writes:

> In case of any unset variable, make ./configure exit with nonzero value;
> an attempt to expand an unset variable is a bug in the script
> (usually a spelling mistake) and those should not pass through
> unnoticed.
> ---
>
> I was also doing set -e support but that has quite a few nontrivial
> things in it. Set -u catches all lagging c++ compilers and so on ;).
> Perhaps later.

This commit has an unintended consequence: on unknown platforms,
configure crashes and doesn't generate Makefile.config at all. In
particular this manifests on Debian/kfreebsd but I imagine there are
other platforms where people are just ignoring the "Notmuch might or
might not build correctly".  I'm tempted to revert the commit for now,
and give ourselves some time after the release of 0.23 to figure out
(and test) the best solution.

d