Chronological Thread 

< Chronological >      < Thread >

---

On 27/05/11 09:56, Lucian Smith wrote:
> While TeLICeMService.hpp is now in the mac and windows include/ directory,
> it is still missing from the linux include/ directory.

Hi Lucian,

I've finally tracked down what happened there; due to a race condition
in the build system where two builds complete at exactly the same time,
the build index for Linux incorrectly referred to the previous version,
and I relied on that index to get the URL I put into the release.

Since there are no source changes, just a correction of an incorrect
URL, I have simply updated the URL to point at the correct file.

Sorry for any inconvenience caused by that.

>
> Would it be too much to ask for the mac binaries to be universal binaries
> instead of i386-only? I personally would like the ppc added; others may
> appreciate x86_64. (Particularly if you don't want to not have people
> have to compile the API themselves on a mac.)

The plan is not to make changes like that for this release; however, the
Auckland Bioengineering Institute is working on setting up build servers
(including a Mac running a 64 bit operating system) that could be used
to produce universal binaries. The suggestion we have had so far is that
supporting x86 or PPC is pointless and we should only be targeting
x86-64 - so if you have a different use case to balance that, it could
help us respond to that other view point when working out requirements.

The downside of universal binaries is that everyone needs to have object
code for all platforms, which is likely to increase the size of the
binaries for everyone on Mac - alternatively, we could make two or three
different builds.

>
> Also, I don't know how these things are usually done on the mac, but the
> provided binaries all assume they are being installed in /usr/local/lib/
> (you can see this with 'otool -L libcevas.dylib', as an example). The
> installation instructions should probably mention this.

You don't have to install them there, however - they work for me
straight out of the SDK directory.

>
> Finally, it would be nice if the final actual install files had different
> names from each other, and if the Windows file was a .zip

That is something we could also look into, but given how close we are to
releasing, not something worth holding up the build for. I'd suggest
opening a tracker item at https://tracker.physiomeproject.org/ (probably
best you do it, so you are CCd and can put the use cases for why you
want that, but if no one does it by Wednesday, I'll try to remember to
do it).

>
> Barring the above issues, I now believe I have gotten the rc to work out
> of the box on a mac, and with that one file copied in by hand on linux.
> My initial attempt at getting it to fully integrate on Windows was not
> successful, but I still have avenues to explore before I admit defeat
> there. So close...

Randall has independently confirmed it works with Java there, and I have
a simple C++ program going on Windows too - any luck since you sent the
e-mail? What exactly is going wrong? Also, what compiler version are you
using (if it is MingW and not MSVC, note that these binaries are only
going to work with MSVC; we are working on MingW binaries, but they
won't be part of this release)?

Best wishes,
Andrew

---

• [cellml-dev] CellML API Release Candidate 4 out, Andrew Miller, 05/24/2011
  • [cellml-dev] [cellml-discussion] CellML API Release Candidate 4 out, Lucian Smith, 05/27/2011
    • [cellml-dev] [cellml-discussion] CellML API Release Candidate 4 out, Andrew Miller, 05/30/2011
      • [cellml-dev] [cellml-discussion] CellML API Release Candidate 4 out, Lucian Smith, 05/30/2011
      • [cellml-dev] [cellml-discussion] CellML API Release Candidate 4 out, Lucian Smith, 05/30/2011

---

Archive powered by MHonArc 2.6.18.