[Bioc-devel] Build/check report news
hpages at fhcrc.org
Mon May 1 21:45:51 CEST 2006
Thanks Kasper for your feedback!
Some of your requests/suggestions are already here:
o A build timestamp:
There are now 2 more timestamps on the report:
1) The time the report itself was generated (t1).
2) The Snapshot Date (t2) i.e. the time the Build System performed
the 'svn up'
on the Bioconductor working copy that it uses to build/check the
t1 - t2 is more or less 9h30 wich is the time required by the Build
System to build
and check all the packages on all the platforms.
The timestamp we always had (t3) and still have is the 'SVN Last
Changed Date' i.e.
the time of the last change in the Bioconductor tree at the time of
We should always see t3 < t2 < t1 or something is broken in the
Build System :-/
With these new timestamps you still don't know when your package was
tested ;-), all you know is that this happened between t2 and t1...
o Other automatic build platforms:
Solaris, Windows and 32-bit Linux builds are back.
o The -Wall flag is back. We've been using it for the 1.8 builds since
It's back in the 1.9 builds too. I will also try to add the
-pedantic flag if this
can lead to even more usefull compiler messages.
Other will come soon:
o More details on the build platform especially compiler versions:
I'm currently working on this. Expect to see some changes before the
end of this week.
o The ability to see the contents of the 00INSTALL: OK, as Seth
suggested, you can force
the installation of your package during the 'R CMD build' step by
having a vignette in it.
Then you can see the output produced by this installation by
following the link in the
BUILD column. Anyway, I'm going to provide the output of the package
This can be done in at least 3 different ways:
1) By providing access to the content of the 00install.out file
during 'R CMD check <pkg>')
2) By providing the output of 'R CMD INSTALL <pkg>'
3) By providing the output of 'install.packages("pkg")' (this could
be followed by
a 'library(pkg)' just to make sure that pkg can be loaded).
I still have to make my mind... May be some of you see some
advantages to one method
over the other. Please let me know.
o I'm also going to start working on the MacOS X builds.
Best, and thanks again for your feedback,
Kasper Daniel Hansen wrote:
> On Apr 25, 2006, at 10:05 PM, Seth Falcon wrote:
>> Hi Kasper,
>> Kasper Daniel Hansen <khansen at stat.berkeley.edu> writes:
>>> In my opinion one of the greatest assets of the BioC build/check
>>> routine is the ability to get your code tested on various platforms,
>>> and that is most important for non-R code (typically C but it may
>>> include other languages).
>>> I would like to see
>>> * Other automatic build platforms like Solaris, Windows and MacOS
>>> X. Perhaps a 32bit Linux as well? (I am sure that this will be the
>>> case in due time, just wanted to mention it :)
>> I'm glad to hear you find the automated builds useful. Note that
>> Herve's announcement for the 1.9 build page is just what we have
>> available _today_. As we did for 1.8, we will have:
>> - Windows
>> - 32 and 64 bit Linux
>> - Solaris
>> And we hope to add at least one OS X build as well. :-)
>>> * More details on the build platform especially compiler versions.
>>> Eg. I have no idea about the difference between SuSE 9.2 and 10.0
>> That's a nice suggestion. I think we will be able to add that easily.
>>> * A build timestamp.
>> I'm not really sure what you want when you say timestamp (I know it
>> sounds obvious, but humor me). The build report says at the top of
>> the page:
>> This is the Multiple platform build/check report for BioC 1.8:
>> - SVN URL: https://hedgehog.fhcrc.org/bioconductor/branches/
>> - SVN Revision: 17679
>> - SVN Last Changed Date: 2006-04-24 22:18:08 -0700 (Mon, 24 Apr
>> This tells you that the report is for a build based on rev 17679 which
>> last changed on 24 April at 22:18. What else can we tell you?
> I have cleaned my glasses and found the above mentioned information.
> That was indeed was I was looking for :)
>>> * The ability to see the contents of the 00INSTALL file which
>>> contains compiler warnings/errors. I would like to see this _even if
>>> the package build successfully_ as there can be some differences in
>>> the warnings the different compilers/platforms yields. This is
>>> something that the old pages lacked as well. At first I looked at the
>>> aCGH package and did indeed see appropriate g++ output, but then I
>>> check the affxparser package and saw none. A bit strange... - indeed
>>> I do not even see the building of help pages? (hmm this difference
>>> also exists on the BioC 1.8 page - why?)
>> We had this same conversation 6 months ago. We'll see what we can do
>> about making the install log available.
>> But all you have to do is add a vignette to your package; it is
>> supposed to have one anyway. Any vignette will do. Then your package
>> will be compiled during R CMD build and you will be able to view the
>> output. :-P
>>> * perhaps it should be considered to use the -Wall flag for gcc
>>> and/or -pendantic?
>> I believe -Wall is already being set for the build, don't know about
> If I look at the build output for aCGH (using g++) and affy (using
> gcc) I do not see the -Wall flag. Also -std=g99 is only set for gcc,
> not g++. I have no idea if that matters.
>>> * If the building process takes too much time, perhaps one only
>>> needs to check code in /src on the various platforms - eg. skip the
>>> generation of help pages?
>> Hmm, that's another interesting idea.
>> Thanks for the comments, hope we can get around to some of the
>> improvements over the next few weeks.
> Thanks, Kasper
>> + seth
>> Bioc-devel at stat.math.ethz.ch mailing list
> Bioc-devel at stat.math.ethz.ch mailing list
E-mail: hpages at fhcrc.org
Phone: (206) 667-5791
Fax: (206) 667-1319
More information about the Bioc-devel