Hi,
we've had binary compiles a while ago, but we were glad
when we could revert that to "normal" compilation a long time ago.
Another observation from years ago was that the linker step
was what took the longest, and even with a precompiled libpwiz.a
this took very long.
Thanks Herve for the observation about normal build times
vs. the "fresh R" case. This might show that, in general,
mzR does build in <40 minutes.
yours,
Steffen
On Wed, 2017-10-11 at 09:39 -0400, Martin Morgan wrote:
On 10/11/2017 08:29 AM, Laurent Gatto wrote:
On 11 October 2017 11:14, Rainer Johannes wrote:
On 11 Oct 2017, at 13:05, Martin Morgan <martin.morgan at roswellp
ark.org> wrote:
On 10/11/2017 03:48 AM, Rainer Johannes wrote:
Hi,
would it be possible to increase the allowed build and check
times for mzR on Windows? Building mzR takes very long due to
the compilation of the included proteowizard (and boost?)
code so there is not much we can do to speed that up.
what's the role of proteowizard in mzR? is it required in its
entirety, or can it be more selectively included? I agree that
mzR is often problematic because of the excessive compilation
time.
mzR uses Rcpp modules to directly call/use the C++ code from
proteowizard to read mzML, mzXML and other MS file formats.
Regarding
selective includes: I believe Steffen Neumann and Laurent Gatto
spent
already a great deal of time to reduce the amount of code that
needs
to be included for mzR to compile/work. I think there is not much
more
that can be done here, unfortunately.
Steffen (he should get the credit for this) has indeed tried to be
as
selective as possible, yes. There might be more to be done, but I
don't
think we could gain anything really substantial.
probably we have been down this road before but is another option to
pre-compile and distribute the Windows binaries?
Martin
On linux and macOS all is fine, but we get TIMEOUT errors on
a regular basis on Windows (I guess because it has to be
compiled for i386 and x86_64). The problem is that a failing
mzR causes all packages depending on it (like MSnbase and
xcms) to fail too.
cheers, jo