pg_upgrade test mods for Windows/Mingw
The attached patch is what I had to do to get pg_upgrade's "make check"
to run on Windows under Mingw. Mostly the changes have to do with
getting paths right between Windows and MSys, or calling generated .bat
files instead of shell scripts.
cheers
andrew
Attachments:
pg_upgrade_test.patchtext/x-patch; name=pg_upgrade_test.patchDownload
diff --git a/contrib/pg_upgrade/test.sh b/contrib/pg_upgrade/test.sh
index 31e30af..fc2304a 100644
--- a/contrib/pg_upgrade/test.sh
+++ b/contrib/pg_upgrade/test.sh
@@ -43,13 +43,23 @@ if [ "$1" = '--install' ]; then
export EXTRA_REGRESS_OPTS
fi
-: ${oldbindir=$bindir}
-
: ${oldsrc=../..}
-oldsrc=`cd "$oldsrc" && pwd`
-newsrc=`cd ../.. && pwd`
-PATH=$bindir:$PATH
+if [ `uname -a | sed 's/.* //'` = Msys ] ; then
+ cp $libdir/libpq.dll $bindir
+ osbindir=`cd $bindir && pwd`
+ bindir=`cd $bindir && pwd -W`
+ oldsrc=`cd "$oldsrc" && pwd -W`
+ newsrc=`cd ../.. && pwd -W`
+else
+ osbindir=$bindir
+ oldsrc=`cd "$oldsrc" && pwd`
+ newsrc=`cd ../.. && pwd`
+fi
+
+: ${oldbindir=$bindir}
+
+PATH=$osbindir:$PATH
export PATH
PGDATA=$temp_root/data
@@ -104,10 +114,19 @@ mv "${PGDATA}" "${PGDATA}.old"
initdb
+if [ `uname -a | sed 's/.* //'` = Msys ] ; then
+ PGDATA=`cd $PGDATA && pwd -W`
+fi
+
pg_upgrade -d "${PGDATA}.old" -D "${PGDATA}" -b "$oldbindir" -B "$bindir"
pg_ctl start -l "$logdir/postmaster2.log" -w
-sh ./analyze_new_cluster.sh
+
+if [ `uname -a | sed 's/.* //'` = Msys ] ; then
+ cmd /c analyze_new_cluster.bat
+else
+ sh ./analyze_new_cluster.sh
+fi
pg_dumpall >"$temp_root"/dump2.sql || pg_dumpall2_status=$?
pg_ctl -m fast stop
if [ -n "$pg_dumpall2_status" ]; then
@@ -115,7 +134,15 @@ if [ -n "$pg_dumpall2_status" ]; then
exit 1
fi
-sh ./delete_old_cluster.sh
+if [ `uname -a | sed 's/.* //'` = Msys ] ; then
+ sed -i -e 's,/,\\,g' -e 's,\\s\\q ,/s/q ,' delete_old_cluster.bat 2>/dev/null
+ chmod +w delete_old_cluster.bat
+ cmd /c delete_old_cluster.bat
+ dos2unix "$temp_root"/dump1.sql >/dev/null
+ dos2unix "$temp_root"/dump2.sql >/dev/null
+else
+ sh ./delete_old_cluster.sh
+fi
if diff -q "$temp_root"/dump1.sql "$temp_root"/dump2.sql; then
echo PASSED
On Sun, Sep 2, 2012 at 11:29 PM, Andrew Dunstan <andrew@dunslane.net> wrote:
The attached patch is what I had to do to get pg_upgrade's "make check" to
run on Windows under Mingw. Mostly the changes have to do with getting
paths right between Windows and MSys, or calling generated .bat files
instead of shell scripts.
When reading shell script code like this
`uname -a | sed 's/.* //'` = Msys
and
sed -i -e 's,/,\\,g' -e 's,\\s\\q ,/s/q ,' delete_old_cluster.bat
2>/dev/null
I find it easier to understand and maintain if the comments also describe
what is the original string format that this pattern-matching expects,
like:
# We expect `uname -a` output like:
# Windows_NT4.0 Msys
and
# We expect lines of the format:
# abc/xyz/def/
# and we convert them to
# abc\xyz\def
BTW, would `uname -o` eliminate the need of pattern matching in the first
snippet? The Wikipedia [1]http://en.wikipedia.org/wiki/Uname article suggests so.
[1]: http://en.wikipedia.org/wiki/Uname
Best regards,
--
Gurjeet Singh
Gurjeet Singh <singh.gurjeet@gmail.com> writes:
[ this needs a comment: ]
`uname -a | sed 's/.* //'` = Msys
Also, how about doing that just once and setting a variable to test
as needed later? Multiple copied-and-pasted instances of line noise
like that are risky.
regards, tom lane
On 09/02/2012 11:53 PM, Gurjeet Singh wrote:
On Sun, Sep 2, 2012 at 11:29 PM, Andrew Dunstan <andrew@dunslane.net
<mailto:andrew@dunslane.net>> wrote:The attached patch is what I had to do to get pg_upgrade's "make
check" to run on Windows under Mingw. Mostly the changes have to
do with getting paths right between Windows and MSys, or calling
generated .bat files instead of shell scripts.When reading shell script code like this
`uname -a | sed 's/.* //'` = Msys
and
sed -i -e 's,/,\\,g' -e 's,\\s\\q ,/s/q ,' delete_old_cluster.bat
2>/dev/nullI find it easier to understand and maintain if the comments also
describe what is the original string format that this
pattern-matching expects, like:# We expect `uname -a` output like:
# Windows_NT4.0 Msysand
# We expect lines of the format:
# abc/xyz/def/
# and we convert them to
# abc\xyz\defBTW, would `uname -o` eliminate the need of pattern matching in the
first snippet? The Wikipedia [1] article suggests so.
Yeah it would. This wasn't intended as a final patch anyway, just as
notice of what I actually had working in case anyone else wanted to try.
cheers
andrew
On 09/03/2012 09:16 AM, Andrew Dunstan wrote:
On 09/02/2012 11:53 PM, Gurjeet Singh wrote:
On Sun, Sep 2, 2012 at 11:29 PM, Andrew Dunstan <andrew@dunslane.net
<mailto:andrew@dunslane.net>> wrote:The attached patch is what I had to do to get pg_upgrade's "make
check" to run on Windows under Mingw. Mostly the changes have to
do with getting paths right between Windows and MSys, or calling
generated .bat files instead of shell scripts.When reading shell script code like this
`uname -a | sed 's/.* //'` = Msys
and
sed -i -e 's,/,\\,g' -e 's,\\s\\q ,/s/q ,' delete_old_cluster.bat
2>/dev/null
BTW, this last one is a hack. pg_upgrade should make sure that it
outputs backslashed paths for rmdir. (In general, the Windows runtime is
quite happy to accept forward-slashed paths, but certain builtin
commands, such as rmdir are not).
cheers
andrew