Hi,
If there is a ~/.cactus/config file, and you also specify options=… to make, how is Cactus supposed to respond? Does it merge the content of the two files, or take one over the other? I suspect that it is merging them. In that case, I think simfactory should complain very loudly if there is a ~/.cactus/config file, as the options you are using will be a mixture of the optionlist configured for the machine and the one in ~/.cactus/config. I'm still dubious about the benefit of having ~/.cactus/config. I think it is used rarely enough by the main developers that it probably isn't tested well, and issues like this tend to get overlooked.
One example of where this causes confusion is if you have MPI = … in ~/.cactus/config, and MPI_DIR = in your optionlist. The MPI thorn complains about MPI being set even though it isn't set in your optionlist, so the error message makes no sense to the user.
I suggest to disable ~/.cactus/config, at least when Simfactory is used.
-erik
On Thu, May 30, 2013 at 10:56 AM, Ian Hinder ian.hinder@aei.mpg.de wrote:
Hi,
If there is a ~/.cactus/config file, and you also specify options=… to make, how is Cactus supposed to respond? Does it merge the content of the two files, or take one over the other? I suspect that it is merging them. In that case, I think simfactory should complain very loudly if there is a ~/.cactus/config file, as the options you are using will be a mixture of the optionlist configured for the machine and the one in ~/.cactus/config. I'm still dubious about the benefit of having ~/.cactus/config. I think it is used rarely enough by the main developers that it probably isn't tested well, and issues like this tend to get overlooked.
One example of where this causes confusion is if you have MPI = … in ~/.cactus/config, and MPI_DIR = in your optionlist. The MPI thorn complains about MPI being set even though it isn't set in your optionlist, so the error message makes no sense to the user.
-- Ian Hinder http://numrel.aei.mpg.de/people/hinder
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
On 30 May 2013, at 17:31, Erik Schnetter schnetter@cct.lsu.edu wrote:
I suggest to disable ~/.cactus/config, at least when Simfactory is used.
Do you mean that SimFactory should abort if it is there, or that we should ask people not to use .cactus/config? People might not remember that they have it there, especially as it is a hidden directory.
-erik
On Thu, May 30, 2013 at 10:56 AM, Ian Hinder ian.hinder@aei.mpg.de wrote: Hi,
If there is a ~/.cactus/config file, and you also specify options=… to make, how is Cactus supposed to respond? Does it merge the content of the two files, or take one over the other? I suspect that it is merging them. In that case, I think simfactory should complain very loudly if there is a ~/.cactus/config file, as the options you are using will be a mixture of the optionlist configured for the machine and the one in ~/.cactus/config. I'm still dubious about the benefit of having ~/.cactus/config. I think it is used rarely enough by the main developers that it probably isn't tested well, and issues like this tend to get overlooked.
One example of where this causes confusion is if you have MPI = … in ~/.cactus/config, and MPI_DIR = in your optionlist. The MPI thorn complains about MPI being set even though it isn't set in your optionlist, so the error message makes no sense to the user.
-- Ian Hinder http://numrel.aei.mpg.de/people/hinder
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
-- Erik Schnetter schnetter@cct.lsu.edu http://www.perimeterinstitute.ca/personal/eschnetter/
We can either remove support for it (ignore it in Cactus, maybe with a warning), or add an option to Cactus that we use from within Simfactory that makes Cactus ignore this file, without a warning.
-erik
On Thu, May 30, 2013 at 11:41 AM, Ian Hinder ian.hinder@aei.mpg.de wrote:
On 30 May 2013, at 17:31, Erik Schnetter schnetter@cct.lsu.edu wrote:
I suggest to disable ~/.cactus/config, at least when Simfactory is used.
Do you mean that SimFactory should abort if it is there, or that we should ask people not to use .cactus/config? People might not remember that they have it there, especially as it is a hidden directory.
-erik
On Thu, May 30, 2013 at 10:56 AM, Ian Hinder ian.hinder@aei.mpg.dewrote:
Hi,
If there is a ~/.cactus/config file, and you also specify options=… to make, how is Cactus supposed to respond? Does it merge the content of the two files, or take one over the other? I suspect that it is merging them. In that case, I think simfactory should complain very loudly if there is a ~/.cactus/config file, as the options you are using will be a mixture of the optionlist configured for the machine and the one in ~/.cactus/config. I'm still dubious about the benefit of having ~/.cactus/config. I think it is used rarely enough by the main developers that it probably isn't tested well, and issues like this tend to get overlooked.
One example of where this causes confusion is if you have MPI = … in ~/.cactus/config, and MPI_DIR = in your optionlist. The MPI thorn complains about MPI being set even though it isn't set in your optionlist, so the error message makes no sense to the user.
-- Ian Hinder http://numrel.aei.mpg.de/people/hinder
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
-- Erik Schnetter schnetter@cct.lsu.edu http://www.perimeterinstitute.ca/personal/eschnetter/
-- Ian Hinder http://numrel.aei.mpg.de/people/hinder
On Thu, May 30, 2013 at 11:44:58AM -0400, Erik Schnetter wrote:
We can either remove support for it (ignore it in Cactus, maybe with a warning), or add an option to Cactus that we use from within Simfactory that makes Cactus ignore this file, without a warning.
I asked on this mailing list earlier whether this file is still used, and only one user replied (Jian Tao). I think we can work with him to make his setup work without.
Any option which ignores the hidden file in one case and not in the other is only adding confusion. This file has been the reason for quite a bit of confusion in the past, including myself before I discontinued to use it. I suggest to remove support for it completely, and complain loudly if such a file exists, to prevent surprises for users.
Frank
Hi,
I still .cactus/config I find it convenient. Could we have .cactus/config be the default if no other config file is specified, and have it ignored if one is?
On 05/30/2013 11:51 AM, Frank Loeffler wrote:
On Thu, May 30, 2013 at 11:44:58AM -0400, Erik Schnetter wrote:
We can either remove support for it (ignore it in Cactus, maybe with a warning), or add an option to Cactus that we use from within Simfactory that makes Cactus ignore this file, without a warning.
I asked on this mailing list earlier whether this file is still used, and only one user replied (Jian Tao). I think we can work with him to make his setup work without.
Any option which ignores the hidden file in one case and not in the other is only adding confusion. This file has been the reason for quite a bit of confusion in the past, including myself before I discontinued to use it. I suggest to remove support for it completely, and complain loudly if such a file exists, to prevent surprises for users.
Frank
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
I use .cactus/config too. What's your suggestion for replacement if someone doesn't want to use simfactory?
Thanks, Bruno.
On 05/30/2013 05:54 PM, Yosef Zlochower wrote:
Hi,
I still .cactus/config I find it convenient. Could we have .cactus/config be the default if no other config file is specified, and have it ignored if one is?
On 05/30/2013 11:51 AM, Frank Loeffler wrote:
On Thu, May 30, 2013 at 11:44:58AM -0400, Erik Schnetter wrote:
We can either remove support for it (ignore it in Cactus, maybe with a warning), or add an option to Cactus that we use from within Simfactory that makes Cactus ignore this file, without a warning.
I asked on this mailing list earlier whether this file is still used, and only one user replied (Jian Tao). I think we can work with him to make his setup work without.
Any option which ignores the hidden file in one case and not in the other is only adding confusion. This file has been the reason for quite a bit of confusion in the past, including myself before I discontinued to use it. I suggest to remove support for it completely, and complain loudly if such a file exists, to prevent surprises for users.
Frank
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
On Thu, May 30, 2013 at 06:09:12PM +0200, Bruno Coutinho Mundim wrote:
I use .cactus/config too. What's your suggestion for replacement if someone doesn't want to use simfactory?
Given that now at least three people spoke up, using .cactus/config, we should no so easily discard it.
What about the following:
- If an option list is specified to 'make', use that - and _not_ .cactus/config. No merging is done, no warning/error is issued. .cactus/config would simply be ignored in that case. This is new. Currently Cactus does some merging here. - If no option list is provided, Cactus uses .cactus/config (if present). No warning or error again, maybe an 'info message'. This doesn't change. - If someone (simfactory?) want to be sure to ignore .cactus/config, it can do so by providing an (possibly empty) option list.
Frank
Although I am still use it, I am not against any proposed changes.
I guess the idea behind this is that explicit is better than implicit, which I agree. As long as any explicit settings replace instead of merging the implicit ones, we shall not have problems.
When one doesn't set any options explicitly, he/she shall be aware that Cactus is doing some magic (.cactus/config is involved)
Regards, Jian
On 05/30/2013 11:51 AM, Frank Loeffler wrote:
On Thu, May 30, 2013 at 11:44:58AM -0400, Erik Schnetter wrote:
We can either remove support for it (ignore it in Cactus, maybe with a warning), or add an option to Cactus that we use from within Simfactory that makes Cactus ignore this file, without a warning.
I asked on this mailing list earlier whether this file is still used, and only one user replied (Jian Tao). I think we can work with him to make his setup work without.
Any option which ignores the hidden file in one case and not in the other is only adding confusion. This file has been the reason for quite a bit of confusion in the past, including myself before I discontinued to use it. I suggest to remove support for it completely, and complain loudly if such a file exists, to prevent surprises for users.
Frank
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
Hello all,
sorry i didn't reply to the last email about the config file...i thought it was directed only to people who use simfactory (which i don't)...
what would the necessary changes be to not use the config file anymore?
best wishes, Vassili
On Thu, May 30, 2013 at 5:51 PM, Frank Loeffler knarf@cct.lsu.edu wrote:
On Thu, May 30, 2013 at 11:44:58AM -0400, Erik Schnetter wrote:
We can either remove support for it (ignore it in Cactus, maybe with a warning), or add an option to Cactus that we use from within Simfactory that makes Cactus ignore this file, without a warning.
I asked on this mailing list earlier whether this file is still used, and only one user replied (Jian Tao). I think we can work with him to make his setup work without.
Any option which ignores the hidden file in one case and not in the other is only adding confusion. This file has been the reason for quite a bit of confusion in the past, including myself before I discontinued to use it. I suggest to remove support for it completely, and complain loudly if such a file exists, to prevent surprises for users.
Frank
Users mailing list Users@einsteintoolkit.org http://lists.einsteintoolkit.org/mailman/listinfo/users
users@lists.einsteintoolkit.org
-
Bruno Coutinho Mundim -
Erik Schnetter -
Frank Loeffler -
Ian Hinder -
Jian Tao -
Vassilios Mewes -
Yosef Zlochower