IcedTea-Web HEAD distcheck broken

[Jiri Vanek]

On 08/25/2011 11:25 PM, Dr Andrew John Hughes wrote:
> On 10:20 Thu 25 Aug     , Jiri Vanek wrote:
>> On 08/25/2011 01:54 AM, Dr Andrew John Hughes wrote:
>>> I haven't done a build for a while, but was inspired to check when replying to
>>> Deepak's OpenBSD e-mail.  It seems distcheck is now broken with IcedTea-Web
>>> HEAD:
>>>
>>> rm: cannot remove `/home/andrew/builder/icedtea-web/icedtea-web-1.2pre/_build/tests.build/netx/unit/net/sourceforge/jnlp/application/application0.jnlp': Permission denied
>>> rm: cannot remove `/home/andrew/builder/icedtea-web/icedtea-web-1.2pre/_build/tests.build/netx/unit/net/sourceforge/jnlp/templates/template0.jnlp': Permission denied
>>> make[1]: *** [clean-netx-unit-tests] Error 1
>>>
>>> Not sure about the branches.
>>
>> distcheck is generated target and is removing permissions from its running directory. chmod -R a-w $(distdir); chmod a+w $(distdir). Because of this permissions eg test reports are not generated.
>
> That's deliberate to catch errors like this.  make distcheck makes the source tree read only.
>
>> I have myself never run this test (and I'm afraid no one have run  this target during this year). Written checking targets are 'make check' for unit test and 'make run-netx-dist-tests' for reproducers tests.
>
> It SHOULD be being run at least before every release.  We always did
> with GNU Classpath and I have with IcedTea7.  It did work when
> IcedTea-Web was first created.  You can check the logs to see the
> changesets I added to fix this.  With something as small as
> IcedTea-Web, there's no reason it shouldn't be run for every change to
> the build.  It ensures it builds with a clean source tree in a
> separate build directory, relative to the source tree, without
> altering the source tree.

Agree... I was afraid you will write something like it. I have checked the logs and i saw your modification. The only one ever done :)

>
>>
>> However,attached patch [RFC] should fix both issues.
>>
>> J.
>
>> --- Makefile.am-old	2011-08-25 08:35:30.100367000 +0200
>> +++ Makefile.am	2011-08-25 10:05:11.780122124 +0200
>> @@ -574,17 +574,18 @@
>>   	rm -f junit-runner-source-files.txt
>>   	rm -rf $(JUNIT_RUNNER_DIR)
>>   	rm -f $(JUNIT_RUNNER_JAR)
>>
>>   clean-netx-unit-tests: clean_tests_reports
>> +	if [ -e $(abs_top_builddir) ] ; then chmod -R a+w $(abs_top_builddir) ; fi;
>>   	rm -f netx-unit-tests-source-files.txt
>>   	rm -rf $(NETX_UNIT_TEST_DIR)
>>   	rm -f stamps/netx-unit-tests-compile.stamp
>>
>>   clean_tests_reports:
>> -	rm -rf  $(TESTS_DIR)/$(REPORT_STYLES_DIRNAME)/
>> -	rm -f  $(TESTS_DIR)/index*.html
>> +	-rm -rf  $(TESTS_DIR)/$(REPORT_STYLES_DIRNAME)/
>> +	-rm -f  $(TESTS_DIR)/index*.html
>>
>>   clean-netx-dist-tests: clean_tests_reports
>>   	rm -f netx-dist-tests-source-files.txt
>>   	rm -rf $(JNLP_TESTS_DIR)
>>   	rm -rf $(JNLP_TESTS_SERVER_DEPLOYDIR)
>
>
> This fix is incorrect.  You shouldn't be applying write permission to the entire build
> directory and that's not the problem anyway.  You're copying files from the source
> tree, but not fixing their permissions.

Well..agree.. It was shot. But as generated code do " chmod -R a-w $(distdir); chmod a+w $(distdir)" then it can not be cleaned!
You know I'm autotolls-lame... Do you have any suggestion here?

Thanx J.

ps chmod -R o+w $(abs_top_builddir) ; Is also wrong? :)