What are the limitations of the Windows Subsystem for Linux? [duplicate]
This question already has an answer here:
Windows Subsystem for Linux (WSL): what can't I do with the Ubuntu application for Microsoft Windows?
4 answers
I played around with the WSL last year, but at the time it was not capable of running the neural network project that I wanted to try on it. Apparently had something to do with it having an outdated kernel or something of the sort.
Since then, there's been a lot of updates to the WSL. What are the limitations of running programs that works on Ubuntu on the WSL? Are they essentially equivalent? Is there some piece of information that I can use to determine this ahead of time?
kernel windows virtualization windows-10 active-directory
edited Jan 24 at 20:27
wjandrea
9,15942363
asked Jan 4 '18 at 20:45
David A. FrenchDavid A. French
87112
marked as duplicate by wjandrea, vidarlo, karel, Charles Green, Kulfy Jan 25 at 16:37
This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.
|
show 1 more comment
This question already has an answer here:
Windows Subsystem for Linux (WSL): what can't I do with the Ubuntu application for Microsoft Windows?
4 answers
I played around with the WSL last year, but at the time it was not capable of running the neural network project that I wanted to try on it. Apparently had something to do with it having an outdated kernel or something of the sort.
Since then, there's been a lot of updates to the WSL. What are the limitations of running programs that works on Ubuntu on the WSL? Are they essentially equivalent? Is there some piece of information that I can use to determine this ahead of time?
kernel windows virtualization windows-10 active-directory
edited Jan 24 at 20:27
wjandrea
9,15942363
asked Jan 4 '18 at 20:45
David A. FrenchDavid A. French
87112
marked as duplicate by wjandrea, vidarlo, karel, Charles Green, Kulfy Jan 25 at 16:37
This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.
@wjandrea The question you linked was asked in July 2018 and this question was originally asked in January 2018.... unless you believe in time travel this ain't a duplicate of the linked question.
– David A. French
Jan 24 at 20:22
1
Duplicate on SE sites means that it's answers are valid for this question as well. It doesn't have to be an exact duplicate, nor does it have to be older than the duplicate. It's merely a way of indicating that the answers to the other question will answer this one as well.
– vidarlo
Jan 24 at 20:25
1
@David Duplicates actually aren't based on time but on quality and popularity. The linked question has more upvotes and more answers. BTW, the questions have a few differences, so maybe they're not exact duplicates, but I'll leave it to you and the community to decide.
– wjandrea
Jan 24 at 20:27
1
Apologies, fair enough. Though, 1) my question was specifically targeted at discussing this stuff in the context of neural networks, which is more specific and not covered by the other question, and 2) while the linked question has more upvotes, the top answer is honestly kind of garbage. My updated answer below contains a lot more information.
– David A. French
Jan 24 at 23:05
1
I have noticed primarily limitations with kernel module availability such as a missing loopback kernel module (which means no support for mounting iso files as a block device). Also direct block device access via /dev is missing for tools like dd
– Jasper
Jan 25 at 23:13
|
show 1 more comment
This question already has an answer here:
Windows Subsystem for Linux (WSL): what can't I do with the Ubuntu application for Microsoft Windows?
4 answers
I played around with the WSL last year, but at the time it was not capable of running the neural network project that I wanted to try on it. Apparently had something to do with it having an outdated kernel or something of the sort.
Since then, there's been a lot of updates to the WSL. What are the limitations of running programs that works on Ubuntu on the WSL? Are they essentially equivalent? Is there some piece of information that I can use to determine this ahead of time?
kernel windows virtualization windows-10 active-directory
edited Jan 24 at 20:27
wjandrea
9,15942363
asked Jan 4 '18 at 20:45
David A. FrenchDavid A. French
87112
This question already has an answer here:
Windows Subsystem for Linux (WSL): what can't I do with the Ubuntu application for Microsoft Windows?
4 answers
I played around with the WSL last year, but at the time it was not capable of running the neural network project that I wanted to try on it. Apparently had something to do with it having an outdated kernel or something of the sort.
Since then, there's been a lot of updates to the WSL. What are the limitations of running programs that works on Ubuntu on the WSL? Are they essentially equivalent? Is there some piece of information that I can use to determine this ahead of time?
This question already has an answer here:
Windows Subsystem for Linux (WSL): what can't I do with the Ubuntu application for Microsoft Windows?
4 answers
kernel windows virtualization windows-10 active-directory
kernel windows virtualization windows-10 active-directory
edited Jan 24 at 20:27
wjandrea
9,15942363
asked Jan 4 '18 at 20:45
David A. FrenchDavid A. French
87112
edited Jan 24 at 20:27
wjandrea
9,15942363
asked Jan 4 '18 at 20:45
David A. FrenchDavid A. French
87112
edited Jan 24 at 20:27
wjandrea
9,15942363
edited Jan 24 at 20:27
wjandrea
9,15942363
edited Jan 24 at 20:27
wjandrea
9,15942363
9,15942363
asked Jan 4 '18 at 20:45
David A. FrenchDavid A. French
87112
asked Jan 4 '18 at 20:45
David A. FrenchDavid A. French
87112
asked Jan 4 '18 at 20:45
David A. FrenchDavid A. French
87112
87112
marked as duplicate by wjandrea, vidarlo, karel, Charles Green, Kulfy Jan 25 at 16:37
This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.
marked as duplicate by wjandrea, vidarlo, karel, Charles Green, Kulfy Jan 25 at 16:37
This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.
@wjandrea The question you linked was asked in July 2018 and this question was originally asked in January 2018.... unless you believe in time travel this ain't a duplicate of the linked question.
– David A. French
Jan 24 at 20:22
1
Duplicate on SE sites means that it's answers are valid for this question as well. It doesn't have to be an exact duplicate, nor does it have to be older than the duplicate. It's merely a way of indicating that the answers to the other question will answer this one as well.
– vidarlo
Jan 24 at 20:25
1
@David Duplicates actually aren't based on time but on quality and popularity. The linked question has more upvotes and more answers. BTW, the questions have a few differences, so maybe they're not exact duplicates, but I'll leave it to you and the community to decide.
– wjandrea
Jan 24 at 20:27
1
Apologies, fair enough. Though, 1) my question was specifically targeted at discussing this stuff in the context of neural networks, which is more specific and not covered by the other question, and 2) while the linked question has more upvotes, the top answer is honestly kind of garbage. My updated answer below contains a lot more information.
– David A. French
Jan 24 at 23:05
1
I have noticed primarily limitations with kernel module availability such as a missing loopback kernel module (which means no support for mounting iso files as a block device). Also direct block device access via /dev is missing for tools like dd
– Jasper
Jan 25 at 23:13
|
show 1 more comment
@wjandrea The question you linked was asked in July 2018 and this question was originally asked in January 2018.... unless you believe in time travel this ain't a duplicate of the linked question.
– David A. French
Jan 24 at 20:22
1
Duplicate on SE sites means that it's answers are valid for this question as well. It doesn't have to be an exact duplicate, nor does it have to be older than the duplicate. It's merely a way of indicating that the answers to the other question will answer this one as well.
– vidarlo
Jan 24 at 20:25
1
@David Duplicates actually aren't based on time but on quality and popularity. The linked question has more upvotes and more answers. BTW, the questions have a few differences, so maybe they're not exact duplicates, but I'll leave it to you and the community to decide.
– wjandrea
Jan 24 at 20:27
1
Apologies, fair enough. Though, 1) my question was specifically targeted at discussing this stuff in the context of neural networks, which is more specific and not covered by the other question, and 2) while the linked question has more upvotes, the top answer is honestly kind of garbage. My updated answer below contains a lot more information.
– David A. French
Jan 24 at 23:05
1
I have noticed primarily limitations with kernel module availability such as a missing loopback kernel module (which means no support for mounting iso files as a block device). Also direct block device access via /dev is missing for tools like dd
– Jasper
Jan 25 at 23:13
@wjandrea The question you linked was asked in July 2018 and this question was originally asked in January 2018.... unless you believe in time travel this ain't a duplicate of the linked question.
– David A. French
Jan 24 at 20:22
@wjandrea The question you linked was asked in July 2018 and this question was originally asked in January 2018.... unless you believe in time travel this ain't a duplicate of the linked question.
– David A. French
Jan 24 at 20:22
1
1
Duplicate on SE sites means that it's answers are valid for this question as well. It doesn't have to be an exact duplicate, nor does it have to be older than the duplicate. It's merely a way of indicating that the answers to the other question will answer this one as well.
– vidarlo
Jan 24 at 20:25
Duplicate on SE sites means that it's answers are valid for this question as well. It doesn't have to be an exact duplicate, nor does it have to be older than the duplicate. It's merely a way of indicating that the answers to the other question will answer this one as well.
– vidarlo
Jan 24 at 20:25
1
1
@David Duplicates actually aren't based on time but on quality and popularity. The linked question has more upvotes and more answers. BTW, the questions have a few differences, so maybe they're not exact duplicates, but I'll leave it to you and the community to decide.
– wjandrea
Jan 24 at 20:27
@David Duplicates actually aren't based on time but on quality and popularity. The linked question has more upvotes and more answers. BTW, the questions have a few differences, so maybe they're not exact duplicates, but I'll leave it to you and the community to decide.
– wjandrea
Jan 24 at 20:27
1
1
Apologies, fair enough. Though, 1) my question was specifically targeted at discussing this stuff in the context of neural networks, which is more specific and not covered by the other question, and 2) while the linked question has more upvotes, the top answer is honestly kind of garbage. My updated answer below contains a lot more information.
– David A. French
Jan 24 at 23:05
Apologies, fair enough. Though, 1) my question was specifically targeted at discussing this stuff in the context of neural networks, which is more specific and not covered by the other question, and 2) while the linked question has more upvotes, the top answer is honestly kind of garbage. My updated answer below contains a lot more information.
– David A. French
Jan 24 at 23:05
1
1
I have noticed primarily limitations with kernel module availability such as a missing loopback kernel module (which means no support for mounting iso files as a block device). Also direct block device access via /dev is missing for tools like dd
– Jasper
Jan 25 at 23:13
I have noticed primarily limitations with kernel module availability such as a missing loopback kernel module (which means no support for mounting iso files as a block device). Also direct block device access via /dev is missing for tools like dd
– Jasper
Jan 25 at 23:13
|
show 1 more comment
2 Answers
2
active
oldest
votes
From the WSL FAQ:
Can I run ALL Linux apps in WSL?
No! WSL is a tool aimed at enabling users who need them to run Bash
and core Linux command-line tools on Windows.
WSL does not aim to support GUI desktops or applications (e.g. Gnome,
KDE, etc.)
Also, even though you will be able to run many popular server
applications (e.g. Redis), we do not recommend WSL for server
scenarios – Microsoft offers a variety of solutions for running
production Ubuntu workloads in Azure, Hyper-V, and Docker.
There is also the official test for the Linux Test Program that provides some insight on pieces of software and if they work or come close to working.
The 2 links above are the official stance. One of the things Microsoft notes is that WSL does not aim to support a graphical interface. Users though have been able to run xfce and unity.
When someone can get unity to work on WSL I doubt there are any limitations. Lots of WSL is unproven ground so to get something to work might be up to the user (ie. in this case you ;) ).
answered Jan 4 '18 at 21:09
RinzwindRinzwind
207k28397528
the limitation is on a deeper level. Since there isn't a Linux kernel, things that use kernel modules, like VirtualBox, won't work. Note; I haven't tested myself so correct me if my assumption is incorrect.
– ravery
Jan 4 '18 at 21:36
Vbox is a bit of a stretch yes. But even then I had the idea it might be possible to do; it would need lots of reworking of the code but I would assume it might be possible to pass all what is needed from the kernel to Windows and have that one handle it. You sort of emulate a kernel on the WSL side that would send to and receive from windows and report that back to vbox. Just rambling though ;)
– Rinzwind
Jan 4 '18 at 21:59
I'd seen the FAQ you pointed to but the neural network tools I'm using are CLI and don not have GUIs or and are not graphical programs. I got an answer from the official WSL github though. WSL does not yet support CUDA or OpenCL which are necessary to use GPU with the neural networks. Even if I could get it to run, CPU only would be basically useless. Guess I'm installing Linux on a partition again!
– David A. French
Jan 6 '18 at 5:39
add a comment |
Since this original question was asked things have settled down a bit with WSL and I'd like to pass along the information.
I can report it is basically possible to do any normal Linux task in WSL that requires only the CPU. That means you can easily run GUI apps, and even entire desktop environments, but it also means that running neural networks is more of an exercise in wasting electricity. (WSL also lacks CUDA drivers though there are CUDA drivers for Ubuntu that could possibly be used if WSL actually could get GPU access)
GUI apps
You can get GUI apps the following way:
All you have to do is:
- Install a Windows X server. I use VcXsrv
- Add the routing information to bash using the following command:
echo "export DISPLAY=localhost:0.0" >> ~/.bashrc
If you run zsh, just add the export display info to your .zshrc file.
If you run zsh, just add the export display info to your .zshrc file.
I solved the wanting to run neural networks issue by... just installing Linux on my desktop, haha. It's much easier that way.
To answer the question more broadly
The most significant limitation of WSL is that the Linux and Windows filesystems are not really compatible. Linux files use a different permissions scheme than Windows files, and when you edit or save a file in the WSL home folder ~/ you save it with metadata that will make Windows just show an empty file in Notepad. There are also some problems with the difference in line endings (LF on Linux vs CRLF on Windows) but this is much less of an issue, especially since Microsoft has enabled VS Code and Notepad to use either line ending.
The Linux filesystem can technically be accessed from Windows Explorer but it's highly inadvisable and you couldn't really do anything with it if you wanted.
edited Jan 25 at 16:24
wjandrea
9,15942363
answered Jan 24 at 19:16
David A. FrenchDavid A. French
87112
regarding your last point about permissions, is that only true for files saved/edited in the Linux home directory? what if you symlink to a Windows directory and edit those files?
– Simon
Jan 31 at 12:08
add a comment |
2 Answers
2
active
oldest
votes
2 Answers
2
active
oldest
votes
active
oldest
votes
active
oldest
votes
From the WSL FAQ:
Can I run ALL Linux apps in WSL?
No! WSL is a tool aimed at enabling users who need them to run Bash
and core Linux command-line tools on Windows.
WSL does not aim to support GUI desktops or applications (e.g. Gnome,
KDE, etc.)
Also, even though you will be able to run many popular server
applications (e.g. Redis), we do not recommend WSL for server
scenarios – Microsoft offers a variety of solutions for running
production Ubuntu workloads in Azure, Hyper-V, and Docker.
There is also the official test for the Linux Test Program that provides some insight on pieces of software and if they work or come close to working.
The 2 links above are the official stance. One of the things Microsoft notes is that WSL does not aim to support a graphical interface. Users though have been able to run xfce and unity.
When someone can get unity to work on WSL I doubt there are any limitations. Lots of WSL is unproven ground so to get something to work might be up to the user (ie. in this case you ;) ).
answered Jan 4 '18 at 21:09
RinzwindRinzwind
207k28397528
the limitation is on a deeper level. Since there isn't a Linux kernel, things that use kernel modules, like VirtualBox, won't work. Note; I haven't tested myself so correct me if my assumption is incorrect.
– ravery
Jan 4 '18 at 21:36
Vbox is a bit of a stretch yes. But even then I had the idea it might be possible to do; it would need lots of reworking of the code but I would assume it might be possible to pass all what is needed from the kernel to Windows and have that one handle it. You sort of emulate a kernel on the WSL side that would send to and receive from windows and report that back to vbox. Just rambling though ;)
– Rinzwind
Jan 4 '18 at 21:59
I'd seen the FAQ you pointed to but the neural network tools I'm using are CLI and don not have GUIs or and are not graphical programs. I got an answer from the official WSL github though. WSL does not yet support CUDA or OpenCL which are necessary to use GPU with the neural networks. Even if I could get it to run, CPU only would be basically useless. Guess I'm installing Linux on a partition again!
– David A. French
Jan 6 '18 at 5:39
add a comment |
From the WSL FAQ:
Can I run ALL Linux apps in WSL?
No! WSL is a tool aimed at enabling users who need them to run Bash
and core Linux command-line tools on Windows.
WSL does not aim to support GUI desktops or applications (e.g. Gnome,
KDE, etc.)
Also, even though you will be able to run many popular server
applications (e.g. Redis), we do not recommend WSL for server
scenarios – Microsoft offers a variety of solutions for running
production Ubuntu workloads in Azure, Hyper-V, and Docker.
There is also the official test for the Linux Test Program that provides some insight on pieces of software and if they work or come close to working.
The 2 links above are the official stance. One of the things Microsoft notes is that WSL does not aim to support a graphical interface. Users though have been able to run xfce and unity.
When someone can get unity to work on WSL I doubt there are any limitations. Lots of WSL is unproven ground so to get something to work might be up to the user (ie. in this case you ;) ).
answered Jan 4 '18 at 21:09
RinzwindRinzwind
207k28397528
the limitation is on a deeper level. Since there isn't a Linux kernel, things that use kernel modules, like VirtualBox, won't work. Note; I haven't tested myself so correct me if my assumption is incorrect.
– ravery
Jan 4 '18 at 21:36
Vbox is a bit of a stretch yes. But even then I had the idea it might be possible to do; it would need lots of reworking of the code but I would assume it might be possible to pass all what is needed from the kernel to Windows and have that one handle it. You sort of emulate a kernel on the WSL side that would send to and receive from windows and report that back to vbox. Just rambling though ;)
– Rinzwind
Jan 4 '18 at 21:59
I'd seen the FAQ you pointed to but the neural network tools I'm using are CLI and don not have GUIs or and are not graphical programs. I got an answer from the official WSL github though. WSL does not yet support CUDA or OpenCL which are necessary to use GPU with the neural networks. Even if I could get it to run, CPU only would be basically useless. Guess I'm installing Linux on a partition again!
– David A. French
Jan 6 '18 at 5:39
add a comment |
From the WSL FAQ:
Can I run ALL Linux apps in WSL?
No! WSL is a tool aimed at enabling users who need them to run Bash
and core Linux command-line tools on Windows.
WSL does not aim to support GUI desktops or applications (e.g. Gnome,
KDE, etc.)
Also, even though you will be able to run many popular server
applications (e.g. Redis), we do not recommend WSL for server
scenarios – Microsoft offers a variety of solutions for running
production Ubuntu workloads in Azure, Hyper-V, and Docker.
There is also the official test for the Linux Test Program that provides some insight on pieces of software and if they work or come close to working.
The 2 links above are the official stance. One of the things Microsoft notes is that WSL does not aim to support a graphical interface. Users though have been able to run xfce and unity.
When someone can get unity to work on WSL I doubt there are any limitations. Lots of WSL is unproven ground so to get something to work might be up to the user (ie. in this case you ;) ).
answered Jan 4 '18 at 21:09
RinzwindRinzwind
207k28397528
From the WSL FAQ:
Can I run ALL Linux apps in WSL?
No! WSL is a tool aimed at enabling users who need them to run Bash
and core Linux command-line tools on Windows.
WSL does not aim to support GUI desktops or applications (e.g. Gnome,
KDE, etc.)
Also, even though you will be able to run many popular server
applications (e.g. Redis), we do not recommend WSL for server
scenarios – Microsoft offers a variety of solutions for running
production Ubuntu workloads in Azure, Hyper-V, and Docker.
There is also the official test for the Linux Test Program that provides some insight on pieces of software and if they work or come close to working.
The 2 links above are the official stance. One of the things Microsoft notes is that WSL does not aim to support a graphical interface. Users though have been able to run xfce and unity.
When someone can get unity to work on WSL I doubt there are any limitations. Lots of WSL is unproven ground so to get something to work might be up to the user (ie. in this case you ;) ).
answered Jan 4 '18 at 21:09
RinzwindRinzwind
207k28397528
answered Jan 4 '18 at 21:09
RinzwindRinzwind
207k28397528
answered Jan 4 '18 at 21:09
RinzwindRinzwind
207k28397528
answered Jan 4 '18 at 21:09
RinzwindRinzwind
207k28397528
207k28397528
the limitation is on a deeper level. Since there isn't a Linux kernel, things that use kernel modules, like VirtualBox, won't work. Note; I haven't tested myself so correct me if my assumption is incorrect.
– ravery
Jan 4 '18 at 21:36
Vbox is a bit of a stretch yes. But even then I had the idea it might be possible to do; it would need lots of reworking of the code but I would assume it might be possible to pass all what is needed from the kernel to Windows and have that one handle it. You sort of emulate a kernel on the WSL side that would send to and receive from windows and report that back to vbox. Just rambling though ;)
– Rinzwind
Jan 4 '18 at 21:59
I'd seen the FAQ you pointed to but the neural network tools I'm using are CLI and don not have GUIs or and are not graphical programs. I got an answer from the official WSL github though. WSL does not yet support CUDA or OpenCL which are necessary to use GPU with the neural networks. Even if I could get it to run, CPU only would be basically useless. Guess I'm installing Linux on a partition again!
– David A. French
Jan 6 '18 at 5:39
add a comment |
the limitation is on a deeper level. Since there isn't a Linux kernel, things that use kernel modules, like VirtualBox, won't work. Note; I haven't tested myself so correct me if my assumption is incorrect.
– ravery
Jan 4 '18 at 21:36
Vbox is a bit of a stretch yes. But even then I had the idea it might be possible to do; it would need lots of reworking of the code but I would assume it might be possible to pass all what is needed from the kernel to Windows and have that one handle it. You sort of emulate a kernel on the WSL side that would send to and receive from windows and report that back to vbox. Just rambling though ;)
– Rinzwind
Jan 4 '18 at 21:59
I'd seen the FAQ you pointed to but the neural network tools I'm using are CLI and don not have GUIs or and are not graphical programs. I got an answer from the official WSL github though. WSL does not yet support CUDA or OpenCL which are necessary to use GPU with the neural networks. Even if I could get it to run, CPU only would be basically useless. Guess I'm installing Linux on a partition again!
– David A. French
Jan 6 '18 at 5:39
the limitation is on a deeper level. Since there isn't a Linux kernel, things that use kernel modules, like VirtualBox, won't work. Note; I haven't tested myself so correct me if my assumption is incorrect.
– ravery
Jan 4 '18 at 21:36
the limitation is on a deeper level. Since there isn't a Linux kernel, things that use kernel modules, like VirtualBox, won't work. Note; I haven't tested myself so correct me if my assumption is incorrect.
– ravery
Jan 4 '18 at 21:36
Vbox is a bit of a stretch yes. But even then I had the idea it might be possible to do; it would need lots of reworking of the code but I would assume it might be possible to pass all what is needed from the kernel to Windows and have that one handle it. You sort of emulate a kernel on the WSL side that would send to and receive from windows and report that back to vbox. Just rambling though ;)
– Rinzwind
Jan 4 '18 at 21:59
Vbox is a bit of a stretch yes. But even then I had the idea it might be possible to do; it would need lots of reworking of the code but I would assume it might be possible to pass all what is needed from the kernel to Windows and have that one handle it. You sort of emulate a kernel on the WSL side that would send to and receive from windows and report that back to vbox. Just rambling though ;)
– Rinzwind
Jan 4 '18 at 21:59
I'd seen the FAQ you pointed to but the neural network tools I'm using are CLI and don not have GUIs or and are not graphical programs. I got an answer from the official WSL github though. WSL does not yet support CUDA or OpenCL which are necessary to use GPU with the neural networks. Even if I could get it to run, CPU only would be basically useless. Guess I'm installing Linux on a partition again!
– David A. French
Jan 6 '18 at 5:39
I'd seen the FAQ you pointed to but the neural network tools I'm using are CLI and don not have GUIs or and are not graphical programs. I got an answer from the official WSL github though. WSL does not yet support CUDA or OpenCL which are necessary to use GPU with the neural networks. Even if I could get it to run, CPU only would be basically useless. Guess I'm installing Linux on a partition again!
– David A. French
Jan 6 '18 at 5:39
add a comment |
Since this original question was asked things have settled down a bit with WSL and I'd like to pass along the information.
I can report it is basically possible to do any normal Linux task in WSL that requires only the CPU. That means you can easily run GUI apps, and even entire desktop environments, but it also means that running neural networks is more of an exercise in wasting electricity. (WSL also lacks CUDA drivers though there are CUDA drivers for Ubuntu that could possibly be used if WSL actually could get GPU access)
GUI apps
You can get GUI apps the following way:
All you have to do is:
- Install a Windows X server. I use VcXsrv
- Add the routing information to bash using the following command:
echo "export DISPLAY=localhost:0.0" >> ~/.bashrc
If you run zsh, just add the export display info to your .zshrc file.
If you run zsh, just add the export display info to your .zshrc file.
I solved the wanting to run neural networks issue by... just installing Linux on my desktop, haha. It's much easier that way.
To answer the question more broadly
The most significant limitation of WSL is that the Linux and Windows filesystems are not really compatible. Linux files use a different permissions scheme than Windows files, and when you edit or save a file in the WSL home folder ~/ you save it with metadata that will make Windows just show an empty file in Notepad. There are also some problems with the difference in line endings (LF on Linux vs CRLF on Windows) but this is much less of an issue, especially since Microsoft has enabled VS Code and Notepad to use either line ending.
The Linux filesystem can technically be accessed from Windows Explorer but it's highly inadvisable and you couldn't really do anything with it if you wanted.
edited Jan 25 at 16:24
wjandrea
9,15942363
answered Jan 24 at 19:16
David A. FrenchDavid A. French
87112
regarding your last point about permissions, is that only true for files saved/edited in the Linux home directory? what if you symlink to a Windows directory and edit those files?
– Simon
Jan 31 at 12:08
add a comment |
Since this original question was asked things have settled down a bit with WSL and I'd like to pass along the information.
I can report it is basically possible to do any normal Linux task in WSL that requires only the CPU. That means you can easily run GUI apps, and even entire desktop environments, but it also means that running neural networks is more of an exercise in wasting electricity. (WSL also lacks CUDA drivers though there are CUDA drivers for Ubuntu that could possibly be used if WSL actually could get GPU access)
GUI apps
You can get GUI apps the following way:
All you have to do is:
- Install a Windows X server. I use VcXsrv
- Add the routing information to bash using the following command:
echo "export DISPLAY=localhost:0.0" >> ~/.bashrc
If you run zsh, just add the export display info to your .zshrc file.
If you run zsh, just add the export display info to your .zshrc file.
I solved the wanting to run neural networks issue by... just installing Linux on my desktop, haha. It's much easier that way.
To answer the question more broadly
The most significant limitation of WSL is that the Linux and Windows filesystems are not really compatible. Linux files use a different permissions scheme than Windows files, and when you edit or save a file in the WSL home folder ~/ you save it with metadata that will make Windows just show an empty file in Notepad. There are also some problems with the difference in line endings (LF on Linux vs CRLF on Windows) but this is much less of an issue, especially since Microsoft has enabled VS Code and Notepad to use either line ending.
The Linux filesystem can technically be accessed from Windows Explorer but it's highly inadvisable and you couldn't really do anything with it if you wanted.
edited Jan 25 at 16:24
wjandrea
9,15942363
answered Jan 24 at 19:16
David A. FrenchDavid A. French
87112
regarding your last point about permissions, is that only true for files saved/edited in the Linux home directory? what if you symlink to a Windows directory and edit those files?
– Simon
Jan 31 at 12:08
add a comment |
Since this original question was asked things have settled down a bit with WSL and I'd like to pass along the information.
I can report it is basically possible to do any normal Linux task in WSL that requires only the CPU. That means you can easily run GUI apps, and even entire desktop environments, but it also means that running neural networks is more of an exercise in wasting electricity. (WSL also lacks CUDA drivers though there are CUDA drivers for Ubuntu that could possibly be used if WSL actually could get GPU access)
GUI apps
You can get GUI apps the following way:
All you have to do is:
- Install a Windows X server. I use VcXsrv
- Add the routing information to bash using the following command:
echo "export DISPLAY=localhost:0.0" >> ~/.bashrc
If you run zsh, just add the export display info to your .zshrc file.
If you run zsh, just add the export display info to your .zshrc file.
I solved the wanting to run neural networks issue by... just installing Linux on my desktop, haha. It's much easier that way.
To answer the question more broadly
The most significant limitation of WSL is that the Linux and Windows filesystems are not really compatible. Linux files use a different permissions scheme than Windows files, and when you edit or save a file in the WSL home folder ~/ you save it with metadata that will make Windows just show an empty file in Notepad. There are also some problems with the difference in line endings (LF on Linux vs CRLF on Windows) but this is much less of an issue, especially since Microsoft has enabled VS Code and Notepad to use either line ending.
The Linux filesystem can technically be accessed from Windows Explorer but it's highly inadvisable and you couldn't really do anything with it if you wanted.
edited Jan 25 at 16:24
wjandrea
9,15942363
answered Jan 24 at 19:16
David A. FrenchDavid A. French
87112
Since this original question was asked things have settled down a bit with WSL and I'd like to pass along the information.
I can report it is basically possible to do any normal Linux task in WSL that requires only the CPU. That means you can easily run GUI apps, and even entire desktop environments, but it also means that running neural networks is more of an exercise in wasting electricity. (WSL also lacks CUDA drivers though there are CUDA drivers for Ubuntu that could possibly be used if WSL actually could get GPU access)
GUI apps
You can get GUI apps the following way:
All you have to do is:
- Install a Windows X server. I use VcXsrv
- Add the routing information to bash using the following command:
echo "export DISPLAY=localhost:0.0" >> ~/.bashrc
If you run zsh, just add the export display info to your .zshrc file.
If you run zsh, just add the export display info to your .zshrc file.
I solved the wanting to run neural networks issue by... just installing Linux on my desktop, haha. It's much easier that way.
To answer the question more broadly
The most significant limitation of WSL is that the Linux and Windows filesystems are not really compatible. Linux files use a different permissions scheme than Windows files, and when you edit or save a file in the WSL home folder ~/ you save it with metadata that will make Windows just show an empty file in Notepad. There are also some problems with the difference in line endings (LF on Linux vs CRLF on Windows) but this is much less of an issue, especially since Microsoft has enabled VS Code and Notepad to use either line ending.
The Linux filesystem can technically be accessed from Windows Explorer but it's highly inadvisable and you couldn't really do anything with it if you wanted.
edited Jan 25 at 16:24
wjandrea
9,15942363
answered Jan 24 at 19:16
David A. FrenchDavid A. French
87112
edited Jan 25 at 16:24
wjandrea
9,15942363
edited Jan 25 at 16:24
wjandrea
9,15942363
edited Jan 25 at 16:24
wjandrea
9,15942363
9,15942363
answered Jan 24 at 19:16
David A. FrenchDavid A. French
87112
answered Jan 24 at 19:16
David A. FrenchDavid A. French
87112
answered Jan 24 at 19:16
David A. FrenchDavid A. French
87112
87112
regarding your last point about permissions, is that only true for files saved/edited in the Linux home directory? what if you symlink to a Windows directory and edit those files?
– Simon
Jan 31 at 12:08
add a comment |
regarding your last point about permissions, is that only true for files saved/edited in the Linux home directory? what if you symlink to a Windows directory and edit those files?
– Simon
Jan 31 at 12:08
regarding your last point about permissions, is that only true for files saved/edited in the Linux home directory? what if you symlink to a Windows directory and edit those files?
– Simon
Jan 31 at 12:08
regarding your last point about permissions, is that only true for files saved/edited in the Linux home directory? what if you symlink to a Windows directory and edit those files?
– Simon
Jan 31 at 12:08
add a comment |
@wjandrea The question you linked was asked in July 2018 and this question was originally asked in January 2018.... unless you believe in time travel this ain't a duplicate of the linked question.
– David A. French
Jan 24 at 20:22
1
Duplicate on SE sites means that it's answers are valid for this question as well. It doesn't have to be an exact duplicate, nor does it have to be older than the duplicate. It's merely a way of indicating that the answers to the other question will answer this one as well.
– vidarlo
Jan 24 at 20:25
1
@David Duplicates actually aren't based on time but on quality and popularity. The linked question has more upvotes and more answers. BTW, the questions have a few differences, so maybe they're not exact duplicates, but I'll leave it to you and the community to decide.
– wjandrea
Jan 24 at 20:27
1
Apologies, fair enough. Though, 1) my question was specifically targeted at discussing this stuff in the context of neural networks, which is more specific and not covered by the other question, and 2) while the linked question has more upvotes, the top answer is honestly kind of garbage. My updated answer below contains a lot more information.
– David A. French
Jan 24 at 23:05
1
I have noticed primarily limitations with kernel module availability such as a missing loopback kernel module (which means no support for mounting iso files as a block device). Also direct block device access via /dev is missing for tools like dd
– Jasper
Jan 25 at 23:13