traverseda
/
cpython
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
cpython/Misc/build.sh

245 lines
8.2 KiB
Bash
Raw Normal View History

Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
#!/bin/sh
## Script to build and test the latest python from svn. It basically
## does this:
## svn up ; ./configure ; make ; make test ; make install ; cd Doc ; make
##
## Logs are kept and rsync'ed to the host. If there are test failure(s),
## information about the failure(s) is mailed.
##
## This script is run on the PSF's machine as user neal via crontab.
##
## Yes, this script would probably be easier in python, but then
## there's a bootstrap problem. What if Python doesn't build?
##
## This script should be fairly clean Bourne shell, ie not too many
## bash-isms. We should try to keep it portable to other Unixes.
## Even though it will probably only run on Linux. I'm sure there are
## several GNU-isms currently (date +%s and readlink).
##
## Perhaps this script should be broken up into 2 (or more) components.
## Building doc is orthogonal to the rest of the python build/test.
##
## FIXME: we should detect test hangs (eg, if they take more than 45 minutes)
## FIXME: we should run valgrind
## FIXME: we should run code coverage
## Utilities invoked in this script include:
get the proper full pathname of the script. spell build properly
2006-01-02 19:22:41 -04:00
## basename, date, dirname, expr, grep, readlink, uname
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
## cksum, make, mutt, rsync, svn
## remember where did we started from
DIR=`dirname $0`
if [ "$DIR" = "" ]; then
DIR="."
fi
## make directory absolute
DIR=`readlink -f $DIR`
get the proper full pathname of the script. spell build properly
2006-01-02 19:22:41 -04:00
FULLPATHNAME="$DIR/`basename $0`"
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
## we want Misc/..
DIR=`dirname $DIR`
## Configurable options
FAILURE_SUBJECT="Python Regression Test Failures"
#FAILURE_MAILTO="YOUR_ACCOUNT@gmail.com"
Ignore the tests which report leaks sporadically. This should cause few enough spurious warnings to send to python-checkins now. If a test leaks that is not in the ignore list (LEAKY_TESTS), all tests which leak will be reported. This includes those in the ignore list. Hopefully that will prompt someone to fix the flakiness.
2006-02-19 14:48:19 -04:00
FAILURE_MAILTO="python-checkins@python.org"
Add an optional address to copy the failure mails to. Detect a conflict in the only file that should have outstanding changes when this script is run. This doesn't matter on the trunk, but does when run on a branch. Trunk always has the date set to today in boilerplate.tex. Each time a release is cut with a different date, a conflict occurs. (We could copy a known good version, but then we would lose changes to this file.)
2007-04-20 02:20:38 -03:00
#FAILURE_CC="optional--uncomment and set to desired address"
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
REMOTE_SYSTEM="neal@dinsdale.python.org"
REMOTE_DIR="/data/ftp.python.org/pub/docs.python.org/dev/"
RESULT_FILE="$DIR/build/index.html"
INSTALL_DIR="/tmp/python-test/local"
RSYNC_OPTS="-aC -e ssh"
Instead of doing a make test, run the regression tests out of the installed copy. This will hopefully catch problems where directories are added under Lib/ but not to Makefile.pre.in. This breaks out the 2 runs of the test suite with and without -O which is also nicer.
2006-06-27 01:09:13 -03:00
# Always run the installed version of Python.
PYTHON=$INSTALL_DIR/bin/python
# Python options and regression test program that should always be run.
More version wibble
2006-08-18 03:14:52 -03:00
REGRTEST_ARGS="-E -tt $INSTALL_DIR/lib/python2.6/test/regrtest.py"
Instead of doing a make test, run the regression tests out of the installed copy. This will hopefully catch problems where directories are added under Lib/ but not to Makefile.pre.in. This breaks out the 2 runs of the test suite with and without -O which is also nicer.
2006-06-27 01:09:13 -03:00
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
REFLOG="build/reflog.txt.out"
Update comments and the skip list, maybe some of these tests don't report failures, we'll see. Skip certain hopeless tests: compiler and logging. compiler will likely always show varying leaks since it doesn't work on a defined set of modules unless -u compiler is specified. But that takes forever (we only run with -u network currently). logging causes hangs when running with -R.
2006-04-12 02:56:00 -03:00
# These tests are not stable and falsely report leaks sometimes.
Ignore the tests which report leaks sporadically. This should cause few enough spurious warnings to send to python-checkins now. If a test leaks that is not in the ignore list (LEAKY_TESTS), all tests which leak will be reported. This includes those in the ignore list. Hopefully that will prompt someone to fix the flakiness.
2006-02-19 14:48:19 -04:00
# The entire leak report will be mailed if any test not in this list leaks.
I don't think we know of any tests that really leak anymore (other than those in leakers).
2006-04-06 04:58:59 -03:00
# Note: test_XXX (none currently) really leak, but are disabled
Update/expand on comments about leaking tests.
2006-03-17 00:52:38 -04:00
# so we don't send spam. Any test which really leaks should only
# be listed here if there are also test cases under Lib/test/leakers.
There really are some tests that are problematic.
2007-04-29 20:53:24 -03:00
LEAKY_TESTS="test_(cmd_line|socket)"
Update comments and the skip list, maybe some of these tests don't report failures, we'll see. Skip certain hopeless tests: compiler and logging. compiler will likely always show varying leaks since it doesn't work on a defined set of modules unless -u compiler is specified. But that takes forever (we only run with -u network currently). logging causes hangs when running with -R.
2006-04-12 02:56:00 -03:00
# Skip these tests altogether when looking for leaks. These tests
# do not need to be stored above in LEAKY_TESTS too.
# test_compiler almost never finishes with the same number of refs
# since it depends on other modules, skip it.
# test_logging causes hangs, skip it.
LEAKY_SKIPS="-x test_compiler test_logging"
Ignore the tests which report leaks sporadically. This should cause few enough spurious warnings to send to python-checkins now. If a test leaks that is not in the ignore list (LEAKY_TESTS), all tests which leak will be reported. This includes those in the ignore list. Hopefully that will prompt someone to fix the flakiness.
2006-02-19 14:48:19 -04:00
Update/expand on comments about leaking tests.
2006-03-17 00:52:38 -04:00
# Change this flag to "yes" for old releases to only update/build the docs.
Support 2.4 (released versions) better. (rsync was not working with adding the *. Only some files/dirs were copied, not everything.)
2006-02-14 04:14:16 -04:00
BUILD_DISABLED="no"
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
## utility functions
current_time() {
date +%s
}
update_status() {
now=`current_time`
time=`expr $now - $3`
echo "<li><a href=\"$2\">$1</a> <font size=\"-1\">($time seconds)</font></li>" >> $RESULT_FILE
}
mail_on_failure() {
if [ "$NUM_FAILURES" != "0" ]; then
Add an optional address to copy the failure mails to. Detect a conflict in the only file that should have outstanding changes when this script is run. This doesn't matter on the trunk, but does when run on a branch. Trunk always has the date set to today in boilerplate.tex. Each time a release is cut with a different date, a conflict occurs. (We could copy a known good version, but then we would lose changes to this file.)
2007-04-20 02:20:38 -03:00
dest=$FAILURE_MAILTO
# FAILURE_CC is optional.
if [ "$FAILURE_CC" != "" ]; then
dest="$dest -c $FAILURE_CC"
fi
mutt -s "$FAILURE_SUBJECT $1 ($NUM_FAILURES)" $dest < $2
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
fi
}
## setup
cd $DIR
mkdir -p build
rm -f $RESULT_FILE build/*.out
rm -rf $INSTALL_DIR
## create results file
TITLE="Automated Python Build Results"
refresh every 12 hours for crazy people like me that keep the page up
2006-01-16 00:37:22 -04:00
echo "<html>" >> $RESULT_FILE
echo " <head>" >> $RESULT_FILE
echo " <title>$TITLE</title>" >> $RESULT_FILE
echo " <meta http-equiv=\"refresh\" content=\"43200\">" >> $RESULT_FILE
echo " </head>" >> $RESULT_FILE
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
echo "<body>" >> $RESULT_FILE
echo "<h2>Automated Python Build Results</h2>" >> $RESULT_FILE
echo "<table>" >> $RESULT_FILE
echo " <tr>" >> $RESULT_FILE
echo " <td>Built on:</td><td>`date`</td>" >> $RESULT_FILE
echo " </tr><tr>" >> $RESULT_FILE
echo " <td>Hostname:</td><td>`uname -n`</td>" >> $RESULT_FILE
echo " </tr><tr>" >> $RESULT_FILE
echo " <td>Platform:</td><td>`uname -srmpo`</td>" >> $RESULT_FILE
echo " </tr>" >> $RESULT_FILE
echo "</table>" >> $RESULT_FILE
echo "<ul>" >> $RESULT_FILE
## update, build, and test
ORIG_CHECKSUM=`cksum $FULLPATHNAME`
F=svn-update.out
start=`current_time`
svn update >& build/$F
err=$?
update_status "Updating" "$F" $start
Support 2.4 (released versions) better. (rsync was not working with adding the *. Only some files/dirs were copied, not everything.)
2006-02-14 04:14:16 -04:00
if [ $err = 0 -a "$BUILD_DISABLED" != "yes" ]; then
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
## FIXME: we should check if this file has changed.
## If it has changed, we should re-run the script to pick up changes.
if [ "$ORIG_CHECKSUM" != "$ORIG_CHECKSUM" ]; then
exec $FULLPATHNAME $@
fi
F=svn-stat.out
start=`current_time`
svn stat >& build/$F
## ignore some of the diffs
NUM_DIFFS=`egrep -vc '^. (@test|db_home|Lib/test/(regrtest\.py|db_home))$' build/$F`
update_status "svn stat ($NUM_DIFFS possibly important diffs)" "$F" $start
F=configure.out
start=`current_time`
./configure --prefix=$INSTALL_DIR --with-pydebug >& build/$F
err=$?
update_status "Configuring" "$F" $start
if [ $err = 0 ]; then
F=make.out
start=`current_time`
make >& build/$F
err=$?
warnings=`grep warning build/$F | egrep -vc "te?mpnam(_r|)' is dangerous,"`
update_status "Building ($warnings warnings)" "$F" $start
if [ $err = 0 ]; then
* Change tabs to spaces * Enable network tests when hunting for leaks * Disable curses and audio tests properly when running -u all
2006-01-02 18:10:10 -04:00
## make install
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
F=make-install.out
start=`current_time`
make install >& build/$F
update_status "Installing" "$F" $start
Mail if there is an installation failure.
2007-06-01 01:56:47 -03:00
mail_on_failure "install" build/$F
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
Hmm, only python2.x is installed, not plain python. Did that change recently?
2006-07-26 01:00:18 -03:00
if [ ! -x $PYTHON ]; then
ln -s ${PYTHON}2.* $PYTHON
fi
* Change tabs to spaces * Enable network tests when hunting for leaks * Disable curses and audio tests properly when running -u all
2006-01-02 18:10:10 -04:00
## make and run basic tests
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
F=make-test.out
start=`current_time`
Instead of doing a make test, run the regression tests out of the installed copy. This will hopefully catch problems where directories are added under Lib/ but not to Makefile.pre.in. This breaks out the 2 runs of the test suite with and without -O which is also nicer.
2006-06-27 01:09:13 -03:00
$PYTHON $REGRTEST_ARGS >& build/$F
Find test failures consistently
2006-02-15 02:07:32 -04:00
NUM_FAILURES=`grep -ic " failed:" build/$F`
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
update_status "Testing basics ($NUM_FAILURES failures)" "$F" $start
Dammit, who the hell spells build with ii? Thanks to Tony Meyer for spotting my boo-boo.
2006-01-02 22:18:01 -04:00
mail_on_failure "basics" build/$F
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
Instead of doing a make test, run the regression tests out of the installed copy. This will hopefully catch problems where directories are added under Lib/ but not to Makefile.pre.in. This breaks out the 2 runs of the test suite with and without -O which is also nicer.
2006-06-27 01:09:13 -03:00
F=make-test-opt.out
start=`current_time`
$PYTHON -O $REGRTEST_ARGS >& build/$F
NUM_FAILURES=`grep -ic " failed:" build/$F`
update_status "Testing opt ($NUM_FAILURES failures)" "$F" $start
mail_on_failure "opt" build/$F
* Change tabs to spaces * Enable network tests when hunting for leaks * Disable curses and audio tests properly when running -u all
2006-01-02 18:10:10 -04:00
## run the tests looking for leaks
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
F=make-test-refleak.out
start=`current_time`
test_cmd_line should not leak any more, ensure an empty reflog file exists if no leaks are found
2006-04-16 22:48:41 -03:00
## ensure that the reflog exists so the grep doesn't fail
touch $REFLOG
Instead of doing a make test, run the regression tests out of the installed copy. This will hopefully catch problems where directories are added under Lib/ but not to Makefile.pre.in. This breaks out the 2 runs of the test suite with and without -O which is also nicer.
2006-06-27 01:09:13 -03:00
$PYTHON $REGRTEST_ARGS -R 4:3:$REFLOG -u network $LEAKY_SKIPS >& build/$F
Don't ever report a failure when the sum of the reference count differences are zero. This should help reduce the false positives. The message about references leaking is maintained to provide as much info as possible rather than simply suppressing the message at the source.
2007-05-11 02:55:15 -03:00
NUM_FAILURES=`egrep -vc "($LEAKY_TESTS|sum=0)" $REFLOG`
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
update_status "Testing refleaks ($NUM_FAILURES failures)" "$F" $start
mail_on_failure "refleak" $REFLOG
* Change tabs to spaces * Enable network tests when hunting for leaks * Disable curses and audio tests properly when running -u all
2006-01-02 18:10:10 -04:00
## now try to run all the tests
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
F=make-testall.out
start=`current_time`
* Change tabs to spaces * Enable network tests when hunting for leaks * Disable curses and audio tests properly when running -u all
2006-01-02 18:10:10 -04:00
## skip curses when running from cron since there's no terminal
## skip sound since it's not setup on the PSF box (/dev/dsp)
Instead of doing a make test, run the regression tests out of the installed copy. This will hopefully catch problems where directories are added under Lib/ but not to Makefile.pre.in. This breaks out the 2 runs of the test suite with and without -O which is also nicer.
2006-06-27 01:09:13 -03:00
$PYTHON $REGRTEST_ARGS -uall -x test_curses test_linuxaudiodev test_ossaudiodev >& build/$F
Find test failures consistently
2006-02-15 02:07:32 -04:00
NUM_FAILURES=`grep -ic " failed:" build/$F`
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
update_status "Testing all except curses and sound ($NUM_FAILURES failures)" "$F" $start
get the proper full pathname of the script. spell build properly
2006-01-02 19:22:41 -04:00
mail_on_failure "all" build/$F
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
fi
fi
fi
## make doc
Instead of doing a make test, run the regression tests out of the installed copy. This will hopefully catch problems where directories are added under Lib/ but not to Makefile.pre.in. This breaks out the 2 runs of the test suite with and without -O which is also nicer.
2006-06-27 01:09:13 -03:00
cd $DIR/Doc
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
F="make-doc.out"
start=`current_time`
Add an optional address to copy the failure mails to. Detect a conflict in the only file that should have outstanding changes when this script is run. This doesn't matter on the trunk, but does when run on a branch. Trunk always has the date set to today in boilerplate.tex. Each time a release is cut with a different date, a conflict occurs. (We could copy a known good version, but then we would lose changes to this file.)
2007-04-20 02:20:38 -03:00
# Doc/commontex/boilerplate.tex is expected to always have an outstanding
# modification for the date. When a release is cut, a conflict occurs.
# This allows us to detect this problem and not try to build the docs
# which will definitely fail with a conflict.
CONFLICTED_FILE=commontex/boilerplate.tex
conflict_count=`grep -c "<<<" $CONFLICTED_FILE`
if [ $conflict_count != 0 ]; then
echo "Conflict detected in $CONFLICTED_FILE. Doc build skipped." > ../build/$F
err=1
else
make >& ../build/$F
err=$?
fi
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
update_status "Making doc" "$F" $start
if [ $err != 0 ]; then
NUM_FAILURES=1
mail_on_failure "doc" ../build/$F
fi
echo "</ul>" >> $RESULT_FILE
echo "</body>" >> $RESULT_FILE
echo "</html>" >> $RESULT_FILE
## copy results
Support 2.4 (released versions) better. (rsync was not working with adding the *. Only some files/dirs were copied, not everything.)
2006-02-14 04:14:16 -04:00
rsync $RSYNC_OPTS html/* $REMOTE_SYSTEM:$REMOTE_DIR
Script to automatically build and test python and doc. The results are copied up to docs.python.org/dev Needs lots more work. Feel free to add code, fixes, or FIXME comments.
2006-01-02 16:07:16 -04:00
cd ../build
rsync $RSYNC_OPTS index.html *.out $REMOTE_SYSTEM:$REMOTE_DIR/results/