On Wed, Aug 3, 2016 at 3:44 PM, 'Bill Hart' via sage-devel <sage-devel@googlegroups.com> wrote: > There are some awful issues with WSL for now. It has a stack limit of 8MB > which means certain programs that expect a >= 16MB stack won't work. ulimit > refuses to increase the stack size.
Thanks for this detailed status report Bill. > Building things can be *incredibly* slow. Not that this shouldn't be a major > issue for now, since it is supposed to run ordinary Ubuntu binaries. No need > to build them specially for WSL. > > The Sage binary tarball takes 25 minutes to untar! > > I found a number of packages that seems to corrupt the .o files during the > build process every now and again, requiring one to delete the .o files by > hand and rebuild them. Of course this doesn't happen on real Ubuntu. > However, it seems to only happen when doing a parallel build. > > We are getting lots of double free and corruption errors on code that works > absolutely fine on real Ubuntu (though this might just be the memory manager > being more sensitive to actual bugs), and spawning seems to result in a Not > Enough Memory error no matter what. > > Moreover, one needs to download and install the Anniversary update, which > takes hours, then one needs to enable developer tools, including the WSL. > That takes ages to install. Then one needs to install Ubuntu, which takes > quite a while in itself. This experience in particular speaks to why Microsoft has insisted that this is currently being positioned as a tool for *developer convenience* only, and not a platform for end-user software distribution. Again, it could be done, but it seems like almost more hassle than the Docker approach, for "casual" users. Great though for developers! > It's Ubuntu 14.04 and many of the packages are quite old, e.g. gcc 4.8. > > So far I am not that impressed. However, it is marked beta, so I hope they > will improve it over time. Yup! > On the other hand, the ordinary Ubuntu-14.04 binary for Sage seems to > actually work. It takes about a minute to start up the first time. > > As a timing comparison I did a Fateman polynomial benchmark on my 2 GHz > Haswell laptop vs a 2.2 GHz K10 Ubuntu server. > > The timing on the laptop was 176s. On the server it took 169s. So that seems > about right. > > A pearce polynomial benchmark took 95 s on the laptop and 85s on the server. > > So as far as I'm concerned compute performance is isomorphic, especially > given that Sage was built from source on the server and I used a generic > Sage binary on the laptop. Great. That's what I would have thought. > Note that one has access to the ordinary Windows file system, which people > were worried about. And 'top' works. Microsoft are definitely on the right > track here, but it still needs more work in my opinion. My guess is there's > a small group within Microsoft responsible for this, and I'm sure they need > the (moral) support of the Open Source community to keep going with this > project, which is almost there. Maybe patience will pay off. To the filesystem yes, but you can't run Windows executables from within the bash shell which is a minus. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
