You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@couchdb.apache.org by Jan Lehnardt <ja...@apache.org> on 2012/04/30 13:24:06 UTC
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
Sorry :)
On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
> Updated Branches:
> refs/heads/master bdcfcca1e -> 9acc34ed9
>
>
> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>
> Revert. Git <3.
>
> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>
>
> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>
> Branch: refs/heads/master
> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
> Parents: bdcfcca1
> Author: Jan Lehnardt <ja...@apache.org>
> Authored: Mon Apr 30 12:23:00 2012 +0100
> Committer: Jan Lehnardt <ja...@apache.org>
> Committed: Mon Apr 30 12:23:00 2012 +0100
>
> ----------------------------------------------------------------------
> etc/couchdb/Makefile.am | 1 -
> 1 files changed, 0 insertions(+), 1 deletions(-)
> ----------------------------------------------------------------------
>
>
> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
> ----------------------------------------------------------------------
> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
> index fe045f9..dd1054b 100644
> --- a/etc/couchdb/Makefile.am
> +++ b/etc/couchdb/Makefile.am
> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
> local_dev.ini: local.ini
> if test ! -f "$@"; then \
> cp $< $@; \
> - chmod +w $@; \
> fi
>
> install-data-hook:
>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini
is writable"""
Posted by Paul Davis <pa...@gmail.com>.
Done.
On Mon, Apr 30, 2012 at 1:12 PM, Jan Lehnardt <ja...@apache.org> wrote:
>
> On Apr 30, 2012, at 18:43 , Paul Davis wrote:
>
>> On Mon, Apr 30, 2012 at 11:35 AM, Jan Lehnardt <ja...@apache.org> wrote:
>>>
>>> On Apr 30, 2012, at 17:26 , Paul Davis wrote:
>>>
>>>> This series of commits confuses me greatly.
>>>
>>> Sorry.
>>>
>>>> There are a lot of reverts
>>>
>>> The sequence is this:
>>>
>>> You added chmod +w.
>>> I reverted chmod +w.
>>> I accidentally added chmod +w, but called it "Revert" (i.e. this commit should never have been pushed)
>>> I reverted the previous commit, effectively removing chmod +w, adding another "Revert" for funsies.
>>>
>>
>> Ahh, gotchya. :D
>>
>>>> and the explanation of it breaking compilation from read only media is
>>>> incorrect.
>>>
>>> Why is the explanation incorrect?
>>>
>>
>> VPATH builds have a read-only source directory and a read/writable
>> build directory. The source directory must remain read only. Any build
>> products can be writable. local_dev.ini is a build product (albeit a
>> very simple build). In our normal non-VPATH builds we rely on
>> local.ini (the "source" of the local_dev.ini build product) to have
>> correct permissions when local_dev.ini is created (because it copies
>> them) so that tests and ./utils/run work correctly.
>>
>> When we run the VPATH build, it creates a read-only version of
>> local.ini which then ends up affecting the file permissions of
>> local_dev.ini which is built in the build directory. This is a bug in
>> that we don't explicitly specify the permissions of local_dev.ini,
>> hence the chmod. This is perfectly acceptable because local_dev.ini is
>> built in the build directory where files can be writable.
>>
>>>> This file should be writable or ./utils/run in a VPATH build can't
>>>> edit its configuration which is incorrect.
>>>
>>> Then we can't use utils/run to start CouchDB to run the JS tests against in make [dist]check. What alternative would you recommend?
>>>
>>
>> We can. It works. You can tell because make distcheck works when
>> local_dev.ini is writable.
>>
>>>
>>>> Also it breaks the tests.
>>>
>>> Yes, as I said in the previous mail, I'm trying find a fix that doesn't require writing to local_dev.ini when running the tests.
>>>
>>
>> I had the same impulse at first. But then I realized that we do in
>> fact want local_dev.ini to be writable regardless of what this test
>> does. We could be super pure and never change the configuration for
>> unit tests in which case I would say keep looking for a solution if
>> you're so inclined. Regardless of whether we find a solution
>> local_dev.ini should remain writable because normal operation requires
>> the last file of the configuration chain to be writable. Making
>> local_dev.ini read-only is forcing make dev to produce a broken
>> development environment.
>>
>> Also, it occurs to me that a solution to make it more pure would be to
>> just append a value for that setting when local_dev.ini is generated.
>> Ie:
>>
>> echo "[stuff]\nkey = $VALUE\n" >> local_dev.ini
>
> Pitch perfect summary, thanks for making your thought process clear!
>
> +1 then on in fact doing the chmod +w. Do you want to have the
> honours, again? :)
>
> Cheers
> Jan
>
>>
>>> Cheers
>>> Jan
>>> --
>>>
>>>
>>>
>>>
>>>>
>>>> On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
>>>>> Sorry :)
>>>>>
>>>>> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>>>>>
>>>>>> Updated Branches:
>>>>>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>>>>>
>>>>>>
>>>>>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>>>>>
>>>>>> Revert. Git <3.
>>>>>>
>>>>>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>>>>>
>>>>>>
>>>>>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>>>>>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>>>>>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>>>>>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>>>>>
>>>>>> Branch: refs/heads/master
>>>>>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>>>>>> Parents: bdcfcca1
>>>>>> Author: Jan Lehnardt <ja...@apache.org>
>>>>>> Authored: Mon Apr 30 12:23:00 2012 +0100
>>>>>> Committer: Jan Lehnardt <ja...@apache.org>
>>>>>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>>>>>
>>>>>> ----------------------------------------------------------------------
>>>>>> etc/couchdb/Makefile.am | 1 -
>>>>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>>>>> ----------------------------------------------------------------------
>>>>>>
>>>>>>
>>>>>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>>>>>> ----------------------------------------------------------------------
>>>>>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>>>>>> index fe045f9..dd1054b 100644
>>>>>> --- a/etc/couchdb/Makefile.am
>>>>>> +++ b/etc/couchdb/Makefile.am
>>>>>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>>>>>> local_dev.ini: local.ini
>>>>>> if test ! -f "$@"; then \
>>>>>> cp $< $@; \
>>>>>> - chmod +w $@; \
>>>>>> fi
>>>>>>
>>>>>> install-data-hook:
>>>>>>
>>>>>
>>>
>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
Posted by Jan Lehnardt <ja...@apache.org>.
On Apr 30, 2012, at 18:43 , Paul Davis wrote:
> On Mon, Apr 30, 2012 at 11:35 AM, Jan Lehnardt <ja...@apache.org> wrote:
>>
>> On Apr 30, 2012, at 17:26 , Paul Davis wrote:
>>
>>> This series of commits confuses me greatly.
>>
>> Sorry.
>>
>>> There are a lot of reverts
>>
>> The sequence is this:
>>
>> You added chmod +w.
>> I reverted chmod +w.
>> I accidentally added chmod +w, but called it "Revert" (i.e. this commit should never have been pushed)
>> I reverted the previous commit, effectively removing chmod +w, adding another "Revert" for funsies.
>>
>
> Ahh, gotchya. :D
>
>>> and the explanation of it breaking compilation from read only media is
>>> incorrect.
>>
>> Why is the explanation incorrect?
>>
>
> VPATH builds have a read-only source directory and a read/writable
> build directory. The source directory must remain read only. Any build
> products can be writable. local_dev.ini is a build product (albeit a
> very simple build). In our normal non-VPATH builds we rely on
> local.ini (the "source" of the local_dev.ini build product) to have
> correct permissions when local_dev.ini is created (because it copies
> them) so that tests and ./utils/run work correctly.
>
> When we run the VPATH build, it creates a read-only version of
> local.ini which then ends up affecting the file permissions of
> local_dev.ini which is built in the build directory. This is a bug in
> that we don't explicitly specify the permissions of local_dev.ini,
> hence the chmod. This is perfectly acceptable because local_dev.ini is
> built in the build directory where files can be writable.
>
>>> This file should be writable or ./utils/run in a VPATH build can't
>>> edit its configuration which is incorrect.
>>
>> Then we can't use utils/run to start CouchDB to run the JS tests against in make [dist]check. What alternative would you recommend?
>>
>
> We can. It works. You can tell because make distcheck works when
> local_dev.ini is writable.
>
>>
>>> Also it breaks the tests.
>>
>> Yes, as I said in the previous mail, I'm trying find a fix that doesn't require writing to local_dev.ini when running the tests.
>>
>
> I had the same impulse at first. But then I realized that we do in
> fact want local_dev.ini to be writable regardless of what this test
> does. We could be super pure and never change the configuration for
> unit tests in which case I would say keep looking for a solution if
> you're so inclined. Regardless of whether we find a solution
> local_dev.ini should remain writable because normal operation requires
> the last file of the configuration chain to be writable. Making
> local_dev.ini read-only is forcing make dev to produce a broken
> development environment.
>
> Also, it occurs to me that a solution to make it more pure would be to
> just append a value for that setting when local_dev.ini is generated.
> Ie:
>
> echo "[stuff]\nkey = $VALUE\n" >> local_dev.ini
Pitch perfect summary, thanks for making your thought process clear!
+1 then on in fact doing the chmod +w. Do you want to have the
honours, again? :)
Cheers
Jan
>
>> Cheers
>> Jan
>> --
>>
>>
>>
>>
>>>
>>> On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
>>>> Sorry :)
>>>>
>>>> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>>>>
>>>>> Updated Branches:
>>>>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>>>>
>>>>>
>>>>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>>>>
>>>>> Revert. Git <3.
>>>>>
>>>>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>>>>
>>>>>
>>>>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>>>>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>>>>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>>>>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>>>>
>>>>> Branch: refs/heads/master
>>>>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>>>>> Parents: bdcfcca1
>>>>> Author: Jan Lehnardt <ja...@apache.org>
>>>>> Authored: Mon Apr 30 12:23:00 2012 +0100
>>>>> Committer: Jan Lehnardt <ja...@apache.org>
>>>>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>>>>
>>>>> ----------------------------------------------------------------------
>>>>> etc/couchdb/Makefile.am | 1 -
>>>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>>>> ----------------------------------------------------------------------
>>>>>
>>>>>
>>>>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>>>>> ----------------------------------------------------------------------
>>>>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>>>>> index fe045f9..dd1054b 100644
>>>>> --- a/etc/couchdb/Makefile.am
>>>>> +++ b/etc/couchdb/Makefile.am
>>>>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>>>>> local_dev.ini: local.ini
>>>>> if test ! -f "$@"; then \
>>>>> cp $< $@; \
>>>>> - chmod +w $@; \
>>>>> fi
>>>>>
>>>>> install-data-hook:
>>>>>
>>>>
>>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
Posted by Jan Lehnardt <ja...@apache.org>.
On Apr 30, 2012, at 18:43 , Paul Davis wrote:
> On Mon, Apr 30, 2012 at 11:35 AM, Jan Lehnardt <ja...@apache.org> wrote:
>>
>> On Apr 30, 2012, at 17:26 , Paul Davis wrote:
>>
>>> This series of commits confuses me greatly.
>>
>> Sorry.
>>
>>> There are a lot of reverts
>>
>> The sequence is this:
>>
>> You added chmod +w.
>> I reverted chmod +w.
>> I accidentally added chmod +w, but called it "Revert" (i.e. this commit should never have been pushed)
>> I reverted the previous commit, effectively removing chmod +w, adding another "Revert" for funsies.
>>
>
> Ahh, gotchya. :D
>
>>> and the explanation of it breaking compilation from read only media is
>>> incorrect.
>>
>> Why is the explanation incorrect?
>>
>
> VPATH builds have a read-only source directory and a read/writable
> build directory. The source directory must remain read only. Any build
> products can be writable. local_dev.ini is a build product (albeit a
> very simple build). In our normal non-VPATH builds we rely on
> local.ini (the "source" of the local_dev.ini build product) to have
> correct permissions when local_dev.ini is created (because it copies
> them) so that tests and ./utils/run work correctly.
>
> When we run the VPATH build, it creates a read-only version of
> local.ini which then ends up affecting the file permissions of
> local_dev.ini which is built in the build directory. This is a bug in
> that we don't explicitly specify the permissions of local_dev.ini,
> hence the chmod. This is perfectly acceptable because local_dev.ini is
> built in the build directory where files can be writable.
>
>>> This file should be writable or ./utils/run in a VPATH build can't
>>> edit its configuration which is incorrect.
>>
>> Then we can't use utils/run to start CouchDB to run the JS tests against in make [dist]check. What alternative would you recommend?
>>
>
> We can. It works. You can tell because make distcheck works when
> local_dev.ini is writable.
>
>>
>>> Also it breaks the tests.
>>
>> Yes, as I said in the previous mail, I'm trying find a fix that doesn't require writing to local_dev.ini when running the tests.
>>
>
> I had the same impulse at first. But then I realized that we do in
> fact want local_dev.ini to be writable regardless of what this test
> does. We could be super pure and never change the configuration for
> unit tests in which case I would say keep looking for a solution if
> you're so inclined. Regardless of whether we find a solution
> local_dev.ini should remain writable because normal operation requires
> the last file of the configuration chain to be writable. Making
> local_dev.ini read-only is forcing make dev to produce a broken
> development environment.
>
> Also, it occurs to me that a solution to make it more pure would be to
> just append a value for that setting when local_dev.ini is generated.
> Ie:
>
> echo "[stuff]\nkey = $VALUE\n" >> local_dev.ini
Pitch perfect summary, thanks for making your thought process clear!
+1 then on in fact doing the chmod +w. Do you want to have the
honours, again? :)
Cheers
Jan
>
>> Cheers
>> Jan
>> --
>>
>>
>>
>>
>>>
>>> On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
>>>> Sorry :)
>>>>
>>>> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>>>>
>>>>> Updated Branches:
>>>>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>>>>
>>>>>
>>>>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>>>>
>>>>> Revert. Git <3.
>>>>>
>>>>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>>>>
>>>>>
>>>>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>>>>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>>>>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>>>>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>>>>
>>>>> Branch: refs/heads/master
>>>>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>>>>> Parents: bdcfcca1
>>>>> Author: Jan Lehnardt <ja...@apache.org>
>>>>> Authored: Mon Apr 30 12:23:00 2012 +0100
>>>>> Committer: Jan Lehnardt <ja...@apache.org>
>>>>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>>>>
>>>>> ----------------------------------------------------------------------
>>>>> etc/couchdb/Makefile.am | 1 -
>>>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>>>> ----------------------------------------------------------------------
>>>>>
>>>>>
>>>>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>>>>> ----------------------------------------------------------------------
>>>>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>>>>> index fe045f9..dd1054b 100644
>>>>> --- a/etc/couchdb/Makefile.am
>>>>> +++ b/etc/couchdb/Makefile.am
>>>>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>>>>> local_dev.ini: local.ini
>>>>> if test ! -f "$@"; then \
>>>>> cp $< $@; \
>>>>> - chmod +w $@; \
>>>>> fi
>>>>>
>>>>> install-data-hook:
>>>>>
>>>>
>>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini
is writable"""
Posted by Paul Davis <pa...@gmail.com>.
On Mon, Apr 30, 2012 at 11:35 AM, Jan Lehnardt <ja...@apache.org> wrote:
>
> On Apr 30, 2012, at 17:26 , Paul Davis wrote:
>
>> This series of commits confuses me greatly.
>
> Sorry.
>
>> There are a lot of reverts
>
> The sequence is this:
>
> You added chmod +w.
> I reverted chmod +w.
> I accidentally added chmod +w, but called it "Revert" (i.e. this commit should never have been pushed)
> I reverted the previous commit, effectively removing chmod +w, adding another "Revert" for funsies.
>
Ahh, gotchya. :D
>> and the explanation of it breaking compilation from read only media is
>> incorrect.
>
> Why is the explanation incorrect?
>
VPATH builds have a read-only source directory and a read/writable
build directory. The source directory must remain read only. Any build
products can be writable. local_dev.ini is a build product (albeit a
very simple build). In our normal non-VPATH builds we rely on
local.ini (the "source" of the local_dev.ini build product) to have
correct permissions when local_dev.ini is created (because it copies
them) so that tests and ./utils/run work correctly.
When we run the VPATH build, it creates a read-only version of
local.ini which then ends up affecting the file permissions of
local_dev.ini which is built in the build directory. This is a bug in
that we don't explicitly specify the permissions of local_dev.ini,
hence the chmod. This is perfectly acceptable because local_dev.ini is
built in the build directory where files can be writable.
>> This file should be writable or ./utils/run in a VPATH build can't
>> edit its configuration which is incorrect.
>
> Then we can't use utils/run to start CouchDB to run the JS tests against in make [dist]check. What alternative would you recommend?
>
We can. It works. You can tell because make distcheck works when
local_dev.ini is writable.
>
>> Also it breaks the tests.
>
> Yes, as I said in the previous mail, I'm trying find a fix that doesn't require writing to local_dev.ini when running the tests.
>
I had the same impulse at first. But then I realized that we do in
fact want local_dev.ini to be writable regardless of what this test
does. We could be super pure and never change the configuration for
unit tests in which case I would say keep looking for a solution if
you're so inclined. Regardless of whether we find a solution
local_dev.ini should remain writable because normal operation requires
the last file of the configuration chain to be writable. Making
local_dev.ini read-only is forcing make dev to produce a broken
development environment.
Also, it occurs to me that a solution to make it more pure would be to
just append a value for that setting when local_dev.ini is generated.
Ie:
echo "[stuff]\nkey = $VALUE\n" >> local_dev.ini
> Cheers
> Jan
> --
>
>
>
>
>>
>> On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
>>> Sorry :)
>>>
>>> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>>>
>>>> Updated Branches:
>>>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>>>
>>>>
>>>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>>>
>>>> Revert. Git <3.
>>>>
>>>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>>>
>>>>
>>>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>>>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>>>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>>>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>>>
>>>> Branch: refs/heads/master
>>>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>>>> Parents: bdcfcca1
>>>> Author: Jan Lehnardt <ja...@apache.org>
>>>> Authored: Mon Apr 30 12:23:00 2012 +0100
>>>> Committer: Jan Lehnardt <ja...@apache.org>
>>>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>>>
>>>> ----------------------------------------------------------------------
>>>> etc/couchdb/Makefile.am | 1 -
>>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>>> ----------------------------------------------------------------------
>>>>
>>>>
>>>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>>>> ----------------------------------------------------------------------
>>>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>>>> index fe045f9..dd1054b 100644
>>>> --- a/etc/couchdb/Makefile.am
>>>> +++ b/etc/couchdb/Makefile.am
>>>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>>>> local_dev.ini: local.ini
>>>> if test ! -f "$@"; then \
>>>> cp $< $@; \
>>>> - chmod +w $@; \
>>>> fi
>>>>
>>>> install-data-hook:
>>>>
>>>
>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
Posted by Jan Lehnardt <ja...@apache.org>.
On Apr 30, 2012, at 17:26 , Paul Davis wrote:
> This series of commits confuses me greatly.
Sorry.
> There are a lot of reverts
The sequence is this:
You added chmod +w.
I reverted chmod +w.
I accidentally added chmod +w, but called it "Revert" (i.e. this commit should never have been pushed)
I reverted the previous commit, effectively removing chmod +w, adding another "Revert" for funsies.
> and the explanation of it breaking compilation from read only media is
> incorrect.
Why is the explanation incorrect?
> This file should be writable or ./utils/run in a VPATH build can't
> edit its configuration which is incorrect.
Then we can't use utils/run to start CouchDB to run the JS tests against in make [dist]check. What alternative would you recommend?
> Also it breaks the tests.
Yes, as I said in the previous mail, I'm trying find a fix that doesn't require writing to local_dev.ini when running the tests.
Cheers
Jan
--
>
> On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
>> Sorry :)
>>
>> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>>
>>> Updated Branches:
>>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>>
>>>
>>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>>
>>> Revert. Git <3.
>>>
>>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>>
>>>
>>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>>
>>> Branch: refs/heads/master
>>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>>> Parents: bdcfcca1
>>> Author: Jan Lehnardt <ja...@apache.org>
>>> Authored: Mon Apr 30 12:23:00 2012 +0100
>>> Committer: Jan Lehnardt <ja...@apache.org>
>>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>>
>>> ----------------------------------------------------------------------
>>> etc/couchdb/Makefile.am | 1 -
>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>> ----------------------------------------------------------------------
>>>
>>>
>>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>>> ----------------------------------------------------------------------
>>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>>> index fe045f9..dd1054b 100644
>>> --- a/etc/couchdb/Makefile.am
>>> +++ b/etc/couchdb/Makefile.am
>>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>>> local_dev.ini: local.ini
>>> if test ! -f "$@"; then \
>>> cp $< $@; \
>>> - chmod +w $@; \
>>> fi
>>>
>>> install-data-hook:
>>>
>>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
Posted by Jan Lehnardt <ja...@apache.org>.
On Apr 30, 2012, at 17:26 , Paul Davis wrote:
> This series of commits confuses me greatly.
Sorry.
> There are a lot of reverts
The sequence is this:
You added chmod +w.
I reverted chmod +w.
I accidentally added chmod +w, but called it "Revert" (i.e. this commit should never have been pushed)
I reverted the previous commit, effectively removing chmod +w, adding another "Revert" for funsies.
> and the explanation of it breaking compilation from read only media is
> incorrect.
Why is the explanation incorrect?
> This file should be writable or ./utils/run in a VPATH build can't
> edit its configuration which is incorrect.
Then we can't use utils/run to start CouchDB to run the JS tests against in make [dist]check. What alternative would you recommend?
> Also it breaks the tests.
Yes, as I said in the previous mail, I'm trying find a fix that doesn't require writing to local_dev.ini when running the tests.
Cheers
Jan
--
>
> On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
>> Sorry :)
>>
>> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>>
>>> Updated Branches:
>>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>>
>>>
>>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>>
>>> Revert. Git <3.
>>>
>>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>>
>>>
>>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>>
>>> Branch: refs/heads/master
>>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>>> Parents: bdcfcca1
>>> Author: Jan Lehnardt <ja...@apache.org>
>>> Authored: Mon Apr 30 12:23:00 2012 +0100
>>> Committer: Jan Lehnardt <ja...@apache.org>
>>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>>
>>> ----------------------------------------------------------------------
>>> etc/couchdb/Makefile.am | 1 -
>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>> ----------------------------------------------------------------------
>>>
>>>
>>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>>> ----------------------------------------------------------------------
>>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>>> index fe045f9..dd1054b 100644
>>> --- a/etc/couchdb/Makefile.am
>>> +++ b/etc/couchdb/Makefile.am
>>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>>> local_dev.ini: local.ini
>>> if test ! -f "$@"; then \
>>> cp $< $@; \
>>> - chmod +w $@; \
>>> fi
>>>
>>> install-data-hook:
>>>
>>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini
is writable"""
Posted by Robert Newson <rn...@apache.org>.
It doesn't need to be writable to fix the tests, Jan tracked the real
problem down. What's happening is we hit ensure_cookie_auth_secret/1
during the cookie_auth.js test, which fails to write to the read-only
.ini file. Instead, we should explicitly set up the cookie auth secret
and pass the non-persistent flag to couch_config. Then the test passes
and all the files can stay read-only.
As to the revert^3, one can only imagine what goes through Jan's mind
sometimes... :)
B.
On 30 April 2012 17:26, Paul Davis <pa...@gmail.com> wrote:
> This series of commits confuses me greatly. There are a lot of reverts
> and the explanation of it breaking compilation from read only media is
> incorrect.
>
> This file should be writable or ./utils/run in a VPATH build can't
> edit its configuration which is incorrect. Also it breaks the tests.
>
> On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
>> Sorry :)
>>
>> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>>
>>> Updated Branches:
>>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>>
>>>
>>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>>
>>> Revert. Git <3.
>>>
>>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>>
>>>
>>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>>
>>> Branch: refs/heads/master
>>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>>> Parents: bdcfcca1
>>> Author: Jan Lehnardt <ja...@apache.org>
>>> Authored: Mon Apr 30 12:23:00 2012 +0100
>>> Committer: Jan Lehnardt <ja...@apache.org>
>>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>>
>>> ----------------------------------------------------------------------
>>> etc/couchdb/Makefile.am | 1 -
>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>> ----------------------------------------------------------------------
>>>
>>>
>>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>>> ----------------------------------------------------------------------
>>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>>> index fe045f9..dd1054b 100644
>>> --- a/etc/couchdb/Makefile.am
>>> +++ b/etc/couchdb/Makefile.am
>>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>>> local_dev.ini: local.ini
>>> if test ! -f "$@"; then \
>>> cp $< $@; \
>>> - chmod +w $@; \
>>> fi
>>>
>>> install-data-hook:
>>>
>>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini
is writable"""
Posted by Robert Newson <rn...@apache.org>.
It doesn't need to be writable to fix the tests, Jan tracked the real
problem down. What's happening is we hit ensure_cookie_auth_secret/1
during the cookie_auth.js test, which fails to write to the read-only
.ini file. Instead, we should explicitly set up the cookie auth secret
and pass the non-persistent flag to couch_config. Then the test passes
and all the files can stay read-only.
As to the revert^3, one can only imagine what goes through Jan's mind
sometimes... :)
B.
On 30 April 2012 17:26, Paul Davis <pa...@gmail.com> wrote:
> This series of commits confuses me greatly. There are a lot of reverts
> and the explanation of it breaking compilation from read only media is
> incorrect.
>
> This file should be writable or ./utils/run in a VPATH build can't
> edit its configuration which is incorrect. Also it breaks the tests.
>
> On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
>> Sorry :)
>>
>> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>>
>>> Updated Branches:
>>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>>
>>>
>>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>>
>>> Revert. Git <3.
>>>
>>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>>
>>>
>>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>>
>>> Branch: refs/heads/master
>>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>>> Parents: bdcfcca1
>>> Author: Jan Lehnardt <ja...@apache.org>
>>> Authored: Mon Apr 30 12:23:00 2012 +0100
>>> Committer: Jan Lehnardt <ja...@apache.org>
>>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>>
>>> ----------------------------------------------------------------------
>>> etc/couchdb/Makefile.am | 1 -
>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>> ----------------------------------------------------------------------
>>>
>>>
>>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>>> ----------------------------------------------------------------------
>>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>>> index fe045f9..dd1054b 100644
>>> --- a/etc/couchdb/Makefile.am
>>> +++ b/etc/couchdb/Makefile.am
>>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>>> local_dev.ini: local.ini
>>> if test ! -f "$@"; then \
>>> cp $< $@; \
>>> - chmod +w $@; \
>>> fi
>>>
>>> install-data-hook:
>>>
>>
Re: git commit: Revert "Revert "Revert "Make sure that local_dev.ini
is writable"""
Posted by Paul Davis <pa...@gmail.com>.
This series of commits confuses me greatly. There are a lot of reverts
and the explanation of it breaking compilation from read only media is
incorrect.
This file should be writable or ./utils/run in a VPATH build can't
edit its configuration which is incorrect. Also it breaks the tests.
On Mon, Apr 30, 2012 at 6:24 AM, Jan Lehnardt <ja...@apache.org> wrote:
> Sorry :)
>
> On Apr 30, 2012, at 12:23 , jan@apache.org wrote:
>
>> Updated Branches:
>> refs/heads/master bdcfcca1e -> 9acc34ed9
>>
>>
>> Revert "Revert "Revert "Make sure that local_dev.ini is writable"""
>>
>> Revert. Git <3.
>>
>> This reverts commit bdcfcca1ec92f379e5376741e5ab103630573ac4.
>>
>>
>> Project: http://git-wip-us.apache.org/repos/asf/couchdb/repo
>> Commit: http://git-wip-us.apache.org/repos/asf/couchdb/commit/9acc34ed
>> Tree: http://git-wip-us.apache.org/repos/asf/couchdb/tree/9acc34ed
>> Diff: http://git-wip-us.apache.org/repos/asf/couchdb/diff/9acc34ed
>>
>> Branch: refs/heads/master
>> Commit: 9acc34ed9ed4315f7cd6f5b8ac45e47f2c0b8cb6
>> Parents: bdcfcca1
>> Author: Jan Lehnardt <ja...@apache.org>
>> Authored: Mon Apr 30 12:23:00 2012 +0100
>> Committer: Jan Lehnardt <ja...@apache.org>
>> Committed: Mon Apr 30 12:23:00 2012 +0100
>>
>> ----------------------------------------------------------------------
>> etc/couchdb/Makefile.am | 1 -
>> 1 files changed, 0 insertions(+), 1 deletions(-)
>> ----------------------------------------------------------------------
>>
>>
>> http://git-wip-us.apache.org/repos/asf/couchdb/blob/9acc34ed/etc/couchdb/Makefile.am
>> ----------------------------------------------------------------------
>> diff --git a/etc/couchdb/Makefile.am b/etc/couchdb/Makefile.am
>> index fe045f9..dd1054b 100644
>> --- a/etc/couchdb/Makefile.am
>> +++ b/etc/couchdb/Makefile.am
>> @@ -73,7 +73,6 @@ default_dev.ini: default.ini.tpl
>> local_dev.ini: local.ini
>> if test ! -f "$@"; then \
>> cp $< $@; \
>> - chmod +w $@; \
>> fi
>>
>> install-data-hook:
>>
>