John H Palmieri wrote:
On Mar 19, 6:21 pm, Minh Nguyen <nguyenmi...@gmail.com> wrote:
Hi David,
On Sat, Mar 20, 2010 at 11:46 AM, Dr. David Kirkby
<david.kir...@onetel.net> wrote:
<SNIP>
Does anyone have any comments on those, or additions?
The guidelines you listed above sound reasonable to me.
I agree, except for part of #4: "The author must provide evidence to
the reviewer". I think we should trust the authors: if they say that
it builds on Solaris, then we should believe that they have tested
it. They shouldn't need to provide anything except a statement.
How about then changing that to:
"The author should provide the output of 'uname -a' on all system where they
have checked the updated package?"
That would ensure
* They fully understand this, and the system they are using is what they think
it is [1]. If you remotely log into a machine to test code, you might not
understand exactly what it is, and may be mistaken and believe it is OS X when
it is Linux.
* A record of where tested. This would be particularly useful for Linux, as
the trac ticket would have a record he distribution(s) where they tested.
I
would also say that the reviewer should test it themselves on as many
platforms as they have access to.
Well, if the reviewer has access to 50 systems (not impossible if they are a
system admin), it would be unreasonable to expect them to test on every one!
Ideally, though not practical in all cases, the reviewer should test on systems
different to what the author did. If the Linux system the author tested was
Debian, it would be good if the reviewer(s) could find a non-Debian Linux system
to test on. So how about changing #4 to:
"4) Does it at build on Linux, OSX and Solaris? The author should provide the
output of 'uname -a' on all system where they have checked the updated package
and place that on the trac ticket.
The reviewer(s) should test on at least one Linux, OS X and Solaris system.
Ideally, these systems should be different to those used by the author. In the
case of the Linux system, a different distribution of Linux would be preferable.
However, in some cases the reviewer may have no option but to test on the same
systems as the author, though they should try to test on different systems if
possible."
Does that revision seem reasonable?
By the way, to facilitate testing on Solaris, we should have a recent
binary build available so people can quickly install a copy on t2.math
(for instance) to which they have write access. It would be great if
there were a tar.gz file advertised in the motd on that machine.
--
John
Yes. I'll create a binary on 't2' for the current release.
Dave
[1] Many years ago, when computer literacy is less than it is today, I heard of
someone going to a job interview and saying they had developed software on a
"wolf" system. It turns it was a VMS system with the hostname of 'wolf'.
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
To unsubscribe from this group, send email to sage-devel+unsubscribegooglegroups.com or
reply to this email with the words "REMOVE ME" as the subject.
