Re: [sage-devel] Native Windows version
> > I wouldn't say I was "deeply insulted" (though I realize I used the > word "insulting" and I regret that), but I am very discouraged. And > not for any good reason either. Just reporting problems as a matter > of fact would not be a problem. It's more the implied notion that a > couple of problems would be a reason not to use it at all. If that > were the case you should throw out all implementations of Sage on all > platforms. > Correction: All implementations of all software on all platforms :-) except maybe the Analytical Engine Proof is in the eating: my guess is that significant numbers of Windows users are actually using Sage on both CoCalc and the Cygwin installer (I imagine we might have numbers on some of this from web analytics and download stats), and so both are a big win for Sage, despite doubtless many remaining hurdles. Thanks to both of you for huge amounts of work on this. -- 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.
Re: [sage-devel] Native Windows version
On Wed, Jan 23, 2019 at 5:10 PM William Stein wrote: > > On Wed, Jan 23, 2019 at 3:36 AM E. Madison Bray wrote: > > > > If I had to run Sage on Windows for some reason today, I would use > > > Docker or VirtualBox.I helped support somebody using the Cygwin > > > version of Sage on Windows at the Sage booth at the Joint Math > > > Meetings last week, and it typically took about 1 minute to start up > > > Sage, which was scary. She also was 100% convinced that copy/paste in > > > the Sage terminal (really cygwin) didn't work, which would make using > > > Sage very painful indeed. It turned out that copy/paste does work if > > > you use the context menu. > > > > I wish you had asked me about this or brought it up with me before > > spreading unfounded FUD based on one experience. I really don't > > appreciate it--it sets a bad example when the creator of Sage says > > "don't use the Windows port of Sage", and is pretty insulting frankly. > > > > What I wrote above was that in my one experience last week helping somebody > with Cygwin Sage. These two statements are just an observation I had while > helping somebody. > > > So these are engineering challenges, but solvable ones, not reasons to > > say "don't use X". > > I didn't write "don't use X". I wrote "I would use Docker or > VirtualBox", and to clarify I literally > meant that "*I* would use..." (I wish I had put I in bold). This is > true, and was I understood that. In my frustration I was probably unclear about this point. I know it was intended as a statement of personal preference. However, you're in a unique position that your words carry more weight than others and should be considered more carefully. I've been on Hacker News and I know how things get taken out of context. "William Stein says don't use the Windows port of Sage", etc. You then went on to justify your supposed personal preference based on one experience you had with someone else using it. If there were some actual bugs you could say so, but instead it came off as "there were some vague problems, so I wouldn't use it". It's also very discouraging because prior to this (or the WSL) a VM was the only way to use Sage on Windows, and your message gave the impression that all the effort that has been put into making a better Windows experience was for nothing if it's somehow a worse experience even than using a VM (it isn't). > in response to somebody who wrote to our mailing list who was > frustrated with using Cygwin, > so it was a fair suggestion. Also, the original poster was > contemplating porting Sage > to windows, so they are probably extremely technically competent. Maybe they're "frustrated with using Cygin" maybe not. All they wrote is "the ugh-worthy Cygwin" which doesn't give any obvious indication as to their experience with it. I used to have the same feeling about it, though I had barely ever used it myself, so it was more of a knee-jerk opinion. Now that I've spent more time with it I've found that it gives a far better and well-considered UNIX-like experience on Windows than most existing alternatives. The only reason I would prefer not to use it are aforementioned technical problems with emulating certain POSIX interfaces (most notably fork()). They may or may not be "extremely technically competent" but possibly not experienced enough with the software at hand to have a good handle on just how much effort such porting would take. I made the same mistake starting out. I do think it's possible though, of course, but it will take more than just one or two people making a burst of contribution. It's going to require a long, sustained (and sustainable) effort. But who knows, maybe I'm wrong. Maybe I like having work/life balance too much. > > The stuff about copy/paste is just nonsense. [...] > > For me, the best measure for usability is listening to users. Sure, but the copy/paste question is kind of outside the scope of Sage, or even Sage on Windows. Terminal emulator environments on any platform are difficult to novice users who don't have much experience with them. I've been teaching Software Carpentry workshops for years, and every time I think I've seen every problem novices have, I'm surprised anew by new workshops. Regardless, you made the suggestion that this was a unique problem to Sage on Windows, and a reason not to use it. Your "solution" was also inaccurate and uninformed. > Anyways, if you have the chance, I would encourage you to run a booth > somewhere and help > random people use Sage on Windows all day for a few days. Or to try > to create some sort It's almost as if people have been doing that! Ever since the first beta, in fact. I personally haven't gotten out there and done that because I don't go to a lot of mathematics conferences, as they're not generally relevant to me personally. But I'm going to one in a few weeks in fact. In the meantime, other people have been doing that, collecting feedback, and
Re: [sage-devel] Native Windows version
>
> Anyways, if you have the chance, I would encourage you to run a booth
> somewhere and help
> random people use Sage on Windows all day for a few days.
Yeah, it's always amazing to me (not just with Windows, though more so
there) how much stuff that becomes automatic to even a moderately
experienced Sage user really does need to be explained and motivated with
brand-new users - who are often new even to computational software if they
didn't have cause to use it in grad school or wherever. I wish we had
time/$ for some real narrated end user studies ("now try to compute 2+2;
tell me what you are doing as you do it") like one of our CS profs always
is touting to our students, I think a lot of interesting stuff would appear.
> Or to try
> to create some sort
> of support community around sage on windows (e.g., a big link right
> when it starts up
> to a discourse forum or something). I would guess that right now the
> canonical support
> forum for sage on windows might be
> https://groups.google.com/forum/#!forum/sage-windows,
> but it has only two posts in the last two years?
>
I think that was really a forum for *development* of Sage on Windows, but
now that the Cygwin + installer solution has been the primary focus, and
within the "normal" Sage development flow, it has become moribund. We do
see a number of Windows questions on ask.sagemath, so I suppose that is a
potential place to look for common issues.
How hard would it be to add some sort of alert window at startup on Windows
to "here is a good place to ask for help if you are a new user"? I don't
know anything about the installer so maybe it would be hard; on OS X
sometimes there is a canonical readme in the dmg and/or a "help" menu but
this is fairly different from that so it might be hard, or hard to put it
in an easily accessible place for newbies.
--
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.
Re: [sage-devel] Native Windows version
On Thu, Jan 24, 2019 at 1:19 AM Isuru Fernando wrote: > > If you want to start porting dependencies to native windows, I'd suggest > getting all dependencies of sage ported to conda. I maintain a lot of sage > dependencies for conda-forge and would love to have more packages ported to > windows and then maybe some day build sage. > > There are lots of packages needing native windows support. Some packages have > windows builds supported, but no conda package. (For eg: ECL). Some packages > just need some small patches to build system to build them. Some packages > need extra care. (For eg: Julian and I discovered today that NTL assumes > sizeof(mp_size_t) is equal to sizeof(long) or sizeof(int) and uses integer > literals with long suffix. For MSVC/MinGW this is not true) I think anyone trying to port Sage's dependencies, in full, to work "natively", no matter how technically competent, will find themselves quickly in over their head just in terms of overhead. Using a POSIX-style toolchain like the MinGW tools is typically easier; I used to maintain builds of some Python projects on MSVC, but it was an enormous extra effort and once I stopped doing it, I don't think anyone else did. Plus dealing with upstream is a huge pain so you have to be prepared to maintain non-trivial patches forever. The true can be said with Cygwin as well sometimes, albeit usually less so. Most Cygwin-related patches tend to be one-or-two liners, though as my efforts in trying to get CPython fully supported on Cygwin show, just porting one project (especially when it's one as non-trivial as CPython) can take up all of one person's time. So I certainly support such efforts and would like to help out. It's just a daunting task. On the conda end of things one thing I've wanted to do is port Cygwin to conda (and vice-versa) so that packages built for Cygwin can be installed with conda. > On Tue, Jan 22, 2019 at 5:15 PM Dima Pasechnik wrote: >> >> On Tue, Jan 22, 2019 at 10:34 PM deSitter wrote: >> > >> > Hello all, >> > >> > I am interested in creating a native Windows port completely independent >> > of the ugh-worthy Cygwin. Does such a project exist? >> >> there are parts of Sage that are really stumbling blocks for such a project, >> GAP one of them. >> You might start with GAP - I'd expect that it might be quite an effort, >> though. >> >> >> > >> > TIA >> > >> > -drl >> > >> > -- >> > 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. >> >> -- >> 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. > > -- > 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. -- 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.
Re: [sage-devel] Native Windows version
If you want to start porting dependencies to native windows, I'd suggest getting all dependencies of sage ported to conda. I maintain a lot of sage dependencies for conda-forge and would love to have more packages ported to windows and then maybe some day build sage. There are lots of packages needing native windows support. Some packages have windows builds supported, but no conda package. (For eg: ECL). Some packages just need some small patches to build system to build them. Some packages need extra care. (For eg: Julian and I discovered today that NTL assumes sizeof(mp_size_t) is equal to sizeof(long) or sizeof(int) and uses integer literals with long suffix. For MSVC/MinGW this is not true) Isuru On Tue, Jan 22, 2019 at 5:15 PM Dima Pasechnik wrote: > On Tue, Jan 22, 2019 at 10:34 PM deSitter > wrote: > > > > Hello all, > > > > I am interested in creating a native Windows port completely independent > of the ugh-worthy Cygwin. Does such a project exist? > > there are parts of Sage that are really stumbling blocks for such a > project, > GAP one of them. > You might start with GAP - I'd expect that it might be quite an effort, > though. > > > > > > TIA > > > > -drl > > > > -- > > 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. > > -- > 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. > -- 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.
Re: [sage-devel] Native Windows version
On Wed, Jan 23, 2019 at 3:36 AM E. Madison Bray wrote: > > If I had to run Sage on Windows for some reason today, I would use > > Docker or VirtualBox.I helped support somebody using the Cygwin > > version of Sage on Windows at the Sage booth at the Joint Math > > Meetings last week, and it typically took about 1 minute to start up > > Sage, which was scary. She also was 100% convinced that copy/paste in > > the Sage terminal (really cygwin) didn't work, which would make using > > Sage very painful indeed. It turned out that copy/paste does work if > > you use the context menu. > > I wish you had asked me about this or brought it up with me before > spreading unfounded FUD based on one experience. I really don't > appreciate it--it sets a bad example when the creator of Sage says > "don't use the Windows port of Sage", and is pretty insulting frankly. > What I wrote above was that in my one experience last week helping somebody with Cygwin Sage. These two statements are just an observation I had while helping somebody. > So these are engineering challenges, but solvable ones, not reasons to > say "don't use X". I didn't write "don't use X". I wrote "I would use Docker or VirtualBox", and to clarify I literally meant that "*I* would use..." (I wish I had put I in bold). This is true, and was in response to somebody who wrote to our mailing list who was frustrated with using Cygwin, so it was a fair suggestion. Also, the original poster was contemplating porting Sage to windows, so they are probably extremely technically competent. > The stuff about copy/paste is just nonsense. [...] For me, the best measure for usability is listening to users. Anyways, if you have the chance, I would encourage you to run a booth somewhere and help random people use Sage on Windows all day for a few days. Or to try to create some sort of support community around sage on windows (e.g., a big link right when it starts up to a discourse forum or something). I would guess that right now the canonical support forum for sage on windows might be https://groups.google.com/forum/#!forum/sage-windows, but it has only two posts in the last two years? In any case, for me at least, if I'm not seeing or answering support questions regularly from actual users, it can be hard to know what is really nonsense. I'm sorry again to have deeply insulted you and appreciate the enormous contributions you've made to Sage, and I'll avoid posting about cygwin publicly after that without consulting with you first. -- William -- 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.
Re: [sage-devel] Native Windows version
On Tue, Jan 22, 2019 at 10:32 PM Jeroen Demeyer wrote: > > On 2019-01-23 00:11, William Stein wrote: > > I helped support somebody using the Cygwin > > version of Sage on Windows at the Sage booth at the Joint Math > > Meetings last week, and it typically took about 1 minute to start up > > Sage, which was scary. > > But mathematical computations (which do not involve syscalls) shouldn't > be slower, right? When doing support for this user, actually running code seemed fine after the 1 minute (?) startup time. Unfortunately, she was mainly asking about how to use init.sage. (Incidentally, there was numerous questions from Sage users, where the answer is: ~/.sage/init.sage. Evidently it is difficult to learn about ~/.sage/init.sage for some reason. Maybe it should be documented more.) I haven't done benchmarks of cygwin+sage in a decade. My totally outdated experience, which I am probably misremembering, was that mathematical computations that do not involve syscalls could easily be much slower on Cygwin that under VirtualBox + Linux on exactly the same hardware (even though both code was compiled using GCC). This is no surprise to me, since memory management is different between the two, and when I last benchmarked Cygwin was 32-bit only, whereas VirtualBox + Linux was 64-bit, which could significantly impact performance.How memory is cached and managed can impact performance of CPU bound code a lot, and the algorithms differ between Linux, Windows, and OS X. I also remember being similarly frustrated that native Sage on OS X was often slower for things I cared about than VirtualBox + Linux + Sage on OS X; but maybe security (which involves memory address randomization) is higher priority than speed for OS X, for good reason. -- William -- 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.
Re: [sage-devel] Native Windows version
> If I had to run Sage on Windows for some reason today, > I would use Docker or VirtualBox. I’ve been using Sage for many years mainly on Windows. First with the Virtual box. This had many disadvantages. On the one hand the VirtualBox was error-prone and often did not work as expected. It also eats up a lot of memory and especially that worksheets are not stored directly on the host system proved to be a disadvantage. I had no joy with it. My experience changed dramatically when the SageMath binaries appeared in the Windows installable form. I never had a problem again, didn’t have to worry about other software and its quirks. And I had my Yupyter notebooks where I can search for them, in my Windows explorer. Regarding performance: in my applications I have never experienced a situation where I assumed to have any advantage if I used a Linux system. For my work it fits perfectly into Windows. I don’t feel the slightest need to interpose another layer again, whatever type it is. For someone who doesn’t care if Cygwin is "ugh-worthy" or "ohh-worthy" some remarks made in this thread do not reflect the experience of a normal Windows user. I am aware that this is only because of extraordinary skills and efforts by E. M. Bray and I would like to express my appreciation for it. Peter -- 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.
Re: [sage-devel] Native Windows version
On Wednesday, January 23, 2019 at 12:12:39 AM UTC+1, William wrote: > > If I had to run Sage on Windows for some reason today, I would use > Docker or VirtualBox. I recently tried out the new Windows 10 Subsystem for LInux and its pretty sleek. I built Sage for shits & giggles and everything worked fine. I would definitely use that nowadays. -- 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.
Re: [sage-devel] Native Windows version
On Wed, Jan 23, 2019 at 7:32 AM Jeroen Demeyer wrote: > > On 2019-01-23 00:11, William Stein wrote: > > I helped support somebody using the Cygwin > > version of Sage on Windows at the Sage booth at the Joint Math > > Meetings last week, and it typically took about 1 minute to start up > > Sage, which was scary. > > But mathematical computations (which do not involve syscalls) shouldn't > be slower, right? Generally not. If they are multi-process (and hence obviously require system calls to start processes or threads) there is some additional overhead. If the multiprocessing algorithm is using a process pool this is of course an O(1) cost. For straight numerical computations on the CPU there's no major performance problem. -- 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.
Re: [sage-devel] Native Windows version
On Wed, Jan 23, 2019 at 12:12 AM William Stein wrote: > > On Tue, Jan 22, 2019 at 2:34 PM deSitter wrote: > > > > Hello all, > > > > I am interested in creating a native Windows port completely independent of > > the ugh-worthy Cygwin. Does such a project exist? > > Not anymore. I came the closest to writing one, but still didn't > really get close (this was a long time ago -- back when Microsoft way > paying REvolution to port R to Windows). It's very difficult, because > numerous Sage dependencies have never been properly ported to Windows. > If I were to try today, I would start by trying to refactor Sage > itself into a sage-core, with minimal dependencies, and other > libraries that depend on sage-core. Once that is done [1], then I > would worry about porting only sage-core to Windows, as a standard > Python library that is Anaconda installable (I would leverage Anaconda > as much as possible [2]). With that done, I would turn to whatever > major modules (e.g., sage-combinat, sage-nt, sage-algebra, etc.) are > of most interest, and try to port them one-by-one to Windows. Some > would not be possible due to dependencies, and others would be easy. > In any case, I never succeeded at a proper port, nor did anybody > else, so don't take my advise regarding Windows too seriously. The above work on restructuring the Sage lib is still very much on my docket, but not likely to happen without more funding just due to the amount of time needed to do it. I do have a plan that's been underway for months for restructuring how sage-the-lib is packaged. > If I had to run Sage on Windows for some reason today, I would use > Docker or VirtualBox.I helped support somebody using the Cygwin > version of Sage on Windows at the Sage booth at the Joint Math > Meetings last week, and it typically took about 1 minute to start up > Sage, which was scary. She also was 100% convinced that copy/paste in > the Sage terminal (really cygwin) didn't work, which would make using > Sage very painful indeed. It turned out that copy/paste does work if > you use the context menu. I wish you had asked me about this or brought it up with me before spreading unfounded FUD based on one experience. I really don't appreciate it--it sets a bad example when the creator of Sage says "don't use the Windows port of Sage", and is pretty insulting frankly. Lots of people are using the Windows version just fine. It's true that startup is slow (albeit faster after the first time), but it's still my biggest priority to improve. It will take some difficult hacks because the main problem is just the large number of DLLs loaded at startup (425 from "$SAGE_LOCAL" alone), and loading DLLs is typically much slower and more heavy-weight than shared libraries on Linux. The main way to improve this would be to make further (large) improvements to lazy-loading capabilities. There's absolutely no reason 425 extension modules need to be loaded just to get a prompt up and a global namespace (most of the global namespace (as in, the list of names) can be "pre-compiled" for example, and its components lazily-loaded upon use). Work like this would of course benefit all platforms. Another more Windows-specific approach, which is common in the Windows world (e.g. for MS Office, and most browsers) is to have kind of a little "fork server" that automatically loads Sage to a usable state in the background at boot, so that when the user actually launches a terminal it's much faster. This is something I've been meaning to try. So these are engineering challenges, but solvable ones, not reasons to say "don't use X". Sage in Docker on Windows works quite nicely provided that: a) You're comfortable using Docker in the first place, which has a not insignificant learning curve. b) You already have a decent terminal emulator for Windows installed, or you have the more modern cmd.exe that has shipped with some more recent Windows updates (and that I think you need some developer tools installed to have; I forget the details). c) You don't care about certain functionality working well like, I don't know, being able to view plots. Sage in a VM in Windows is a terrible experience for anyone who just wants to use Sage on Windows. The stuff about copy/paste is just nonsense. I'm not sure what the problem there was. It is *slightly* unintuitive because it's not CTRL+C/CTRL+V. Rather, it works more like many *nix terminal emulators in that merely selecting text copies it. Paste is done with a middle-mouse click (if you have a middle button) or with Shift+Ins. Why Shift+Ins I'm not sure, but that's common in terminal emulators on Windows, including cmd.exe. I think CTRL+V might interfere with something else. Regardless, the correct shortcut key is listed in the context menu. -- 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
Re: [sage-devel] Native Windows version
On Tue, Jan 22, 2019 at 11:34 PM deSitter wrote: > > Hello all, > > I am interested in creating a native Windows port completely independent of > the ugh-worthy Cygwin. Does such a project exist? > > TIA > > -drl This is something I had planned to do from the start. It was never my preference to use Cygwin, and instead was going to focus on just using the more minimal MSYS libs. However, porting all of Sage's *dependencies* and especially the other CAS's such as GAP and Singular to work natively is extremely non-trivial (as is porting some other software like cysignals, though Vincent Klein has made great progress on that). I looked early on into some of the dependencies and what it would take to port them. I think GAP, for example, would not be *that* hard. But it's still a non-trivial amount of work. One of the biggest hurdles is simply replacing all the uses of fork() throughout Sage and various dependencies with the appropriate alternative, which may vary from case to case depending on how it's being used. If you would like to find funding to pay my salary for the next two to three years I'd be happy to do it. P.S. I would not characterize Cygwin as "ugh-worthy". Cygwin is a tour-de-force of engineering and is the only reason making Sage work on Windows was feasible at all, as well as provide a usable UNIX shell experience along with it. My only serious motivation for moving away from Cygwin would be the performance and stability problems surrounding fork(). It's remarkable that Cygwin is even able to emulate fork() at all, but it's once of its shakiest areas. It would probably help if MSFT open-sourced parts of the WSL, or at least gave documentation and guidance on how to use the undocumented RtlCloneUserProcess interface that is known to exist in the NT kernel but that nobody really knows how to use: https://www.cygwin.com/ml/cygwin/2018-01/msg00201.html -- 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.
Re: [sage-devel] Native Windows version
On 2019-01-23 00:11, William Stein wrote: I helped support somebody using the Cygwin version of Sage on Windows at the Sage booth at the Joint Math Meetings last week, and it typically took about 1 minute to start up Sage, which was scary. But mathematical computations (which do not involve syscalls) shouldn't be slower, right? -- 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.
Re: [sage-devel] Native Windows version
On Tue, Jan 22, 2019 at 10:34 PM deSitter wrote: > > Hello all, > > I am interested in creating a native Windows port completely independent of > the ugh-worthy Cygwin. Does such a project exist? there are parts of Sage that are really stumbling blocks for such a project, GAP one of them. You might start with GAP - I'd expect that it might be quite an effort, though. > > TIA > > -drl > > -- > 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. -- 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.
Re: [sage-devel] Native Windows version
On Tue, Jan 22, 2019 at 2:34 PM deSitter wrote: > > Hello all, > > I am interested in creating a native Windows port completely independent of > the ugh-worthy Cygwin. Does such a project exist? Not anymore. I came the closest to writing one, but still didn't really get close (this was a long time ago -- back when Microsoft way paying REvolution to port R to Windows). It's very difficult, because numerous Sage dependencies have never been properly ported to Windows. If I were to try today, I would start by trying to refactor Sage itself into a sage-core, with minimal dependencies, and other libraries that depend on sage-core. Once that is done [1], then I would worry about porting only sage-core to Windows, as a standard Python library that is Anaconda installable (I would leverage Anaconda as much as possible [2]). With that done, I would turn to whatever major modules (e.g., sage-combinat, sage-nt, sage-algebra, etc.) are of most interest, and try to port them one-by-one to Windows. Some would not be possible due to dependencies, and others would be easy. In any case, I never succeeded at a proper port, nor did anybody else, so don't take my advise regarding Windows too seriously. If I had to run Sage on Windows for some reason today, I would use Docker or VirtualBox.I helped support somebody using the Cygwin version of Sage on Windows at the Sage booth at the Joint Math Meetings last week, and it typically took about 1 minute to start up Sage, which was scary. She also was 100% convinced that copy/paste in the Sage terminal (really cygwin) didn't work, which would make using Sage very painful indeed. It turned out that copy/paste does work if you use the context menu. -- William [1] I realize that maybe Sage is such a complicated Python library at this point (million lines of code), with so many outstanding "needs review" tickets on trac (always over 200!) that such a refactor would be brutally difficult for any single human being, and very invasive to Sage development. It's probably totally unrealistic. [2] When I was trying hard to do the Windows port of Sage in 2007 (?), I was very, very frustrated because the Anaconda (or whatever it was called) was all a closed source product of Enthought, and we couldn't use it for our Sage port. I pointed this out during my invited Scipy talk, and it made the people who invited me pretty angry at me. In any case, Anaconda has since been very generously made open source. I would make step 0 of porting sage to windows be to fully understand and appreciate what Anaconda on Windows is currently, and very likely build on that. The build system for something complicated like Sage on windows is really important, and the current Sage build system doesn't make sense to use natively with Windows. > > TIA > > -drl > > -- > 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. -- William (http://wstein.org) -- 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.
