On Wed, Apr 14, 2021 at 2:44 AM Floh <flo...@gmail.com> wrote:
> > Is there some reason you don't want to add the emscripten directory to
> you PATH? Is that somehow worse than setting EM_CONFIG and then parsing it?
>
> Because there isn't a single "emscripten directory", but one emscripten
> directory per "workspace directory". I might cd to a different workspace
> with its own embedded emscripten SDK installation at any time in the same
> terminal session. With my current setup, the correct emscripten SDK is
> automatically selected depending on the directory the user is currently in
> (and then when running cmake to generate the build scripts, cmake will
> "bake" the absolute paths into the generated build scripts based on a cmake
> option which points to EMSCRIPTEN_ROOT (which however doesn't exist as
> environment variable)).
>
> > You say you are setting `EM_CONFIG` in your environment [...] Is there
> some reason you couldn't alternatively set `EMSCRIPTEN_ROOT`?
>
> EM_CONFIG is not set persistently in the user environment, it's only set
> temporary inside a python script (as "os.environ['EM_CONFIG'] = ...")
> before the .emscripten file is evaluated, and only because of this one line
> in the .emscripten file:
>
> emsdk_path = os.path.dirname(os.environ.get('EM_CONFIG')).replace('\\',
> '/')
>
> The EM_CONFIG environment variable doesn't survive execution of this
> python script however.
>
> > to point to emscripten directly via `EMSCRIPTEN_ROOT` ...
>
> Isn't this a chicken-egg problem? The EMSCRIPTEN_ROOT variable isn't set
> anywhere on my machine, except after running the platform-specific
> emsdk_env.xxx scripts (which I don't usually do, except when I need emcc in
> the path for some reason). I'd need to run this script (or 'emsdk
> construct_env') of a specific emscripten SDK installation from inside
> python. But running emsdk from within Python which sets environment
> variables doesn't pass this new environment up to the Python process where
> the environment variables would be needed for invoking cmake (AFAIK
> environment variables can only be inherited from parent to child processes,
> but not the other way around).
>
> All in all, the way the current Emscripten SDK works when activated as
> "--embedded" is pretty much perfect because everything already works
> without anything in the PATH or in environment variables. The only wonky
> part is how to get from the SDK installation directory (where 'emsdk' is
> located - which is under my control) to the 'compiler toolchain directory'
> which is needed by cmake (e.g. "[emsdk_dir]/upstream/emscripten"). I
> consider the current way to evaluate .emscripten inside python a hack, but
> it's a reasonably simple hack that works ;)
>
> Any alternative should be simpler then that though ("emsdk construct_env"
> does a lot more things then just setting up 'EMSCRIPTEN_ROOT', and as I
> described above, doesn't really work for my use case because the modified
> environment is "lost" in a child process).
>
>
Thanks for all the details. I feel confident we can come up with a nicer
version of your current hack if we work together. I will open an emsdk
issue for this and maybe reach out to you on discord (are you there?) to
chat about the details.
cheers,
sam
Phew :)
> -Floh.
>
> On Wednesday, 14 April 2021 at 08:16:57 UTC+2 s...@google.com wrote:
>
>> On Tue, Apr 13, 2021 at 1:27 PM Floh <flo...@gmail.com> wrote:
>>
>>> Yeah, I never used python literals in EM_CONFIG, misunderstanding on my
>>> part :)
>>>
>>> About the other thing:
>>>
>>> > a) Use `which emcc` to find where emscripten is
>>>
>>> In my case, emcc is not in the path because this is a local installation
>>> (via "emsdk activate --embedded") without any traces in the environment
>>> (I'm setting EM_CONFIG only temporary in the "python environment" before
>>> evaluating .emscripten as a workaround.
>>>
>>>
>> Is there some reason you don't want to add the emscripten directory to
>> you PATH? Is that somehow worse than setting EM_CONFIG and then parsing it?
>>
>> > b) Have emsdk set `EMSCRIPTEN_ROOT` directly in the environment rather
>>> than in the config file.
>>>
>>> ...this also isn't an option for the same reason as above, there is no
>>> "global emscripten installation" at all, and there may be multiple
>>> installations of the emscripten SDK with different versions side by side.
>>>
>>>
>> I'm not sure I understand. You say you are setting `EM_CONFIG` in your
>> environment according to your needs, so that there is not one global
>> `EM_CONFIG`. Is there some reason you couldn't alternatively set
>> `EMSCRIPTEN_ROOT`?
>>
>> I don't see how either of these imply some "global emscripten
>> installation". In both cases the current environment determins the
>> installation to use, no?. In other words, would it be possible (in your
>> setup) to point to emscripten directly via `EMSCRIPTEN_ROOT` rather than
>> pointing at a config file via `EM_CONFIG` (and then parsing the config file
>> to find EMSCRIPTEN_ROOT)? (it would then be up to the emcc executable to
>> find its own config file).
>>
>>
>>
>>> Cmake takes are of making all paths (compiler tools, header search paths
>>> etc...) absolute starting at EMSCRIPTEN_ROOT (which cmake doesn't read from
>>> an environement variable but gets as command line argument, like cmake
>>> -DEMSCRIPTEN_ROOT=xxx.
>>>
>>> However as long as EMSCRIPTEN_ROOT is always is at
>>> "/upstream/emscripten" under the SDK's installation directory, I think I
>>> don't need to "parse" .emscripten any longer (IFIR in the past this wasn't
>>> true, it was something like "/[version]/emscripten", and this was why I had
>>> to peek into .emscripten in the first place.
>>>
>>> Some other alternatives which would work for me:
>>>
>>> - turn .emscripten into a plain text file with simple "key = value"
>>> lines which is generated at SDK installation time (no more python magic)
>>>
>>> - ...or add a subcommand to the emsdk tool which prints the emscripten
>>> root directory to stdout (which I could then call in my cmake wrapper
>>> scripts)
>>>
>>
>>> I basically need a way to get the absolute directory where emcc is
>>> located if all I know is the absolute directory where the emsdk tool is
>>> located.
>>>
>>
>>> Cheers!
>>> -Andre.
>>>
>>> On Tuesday, 13 April 2021 at 19:08:28 UTC+2 s...@google.com wrote:
>>>
>>>> On Tue, Apr 13, 2021 at 3:46 AM Floh <flo...@gmail.com> wrote:
>>>>
>>>>> I guess this was released in 2.0.17 yesterday (according to the
>>>>> release notes at least)? Happy to report that nothing broke on my side ;)
>>>>>
>>>>> (I'm extracting the EMSCRIPTEN_ROOT value for use in a cmake toolchain
>>>>> file by evaluating the .emscripten config file, and this still seems to
>>>>> work, I'm still setting an EM_CONFIG environment variable before running
>>>>> the .emscripten file through python, but I guess this isn't needed anymore
>>>>> now, but it doesn't hurt either).
>>>>>
>>>>
>>>> Its really a separate issue but I'd love to move folks away from
>>>> parsing `.emscripten` using non-emscripten tools. Do you think we can
>>>> find a way to move you away from doing that?
>>>>
>>>> How about:
>>>>
>>>> a) Use `which emcc` to find where emscripten is
>>>> b) Have emsdk set `EMSCRIPTEN_ROOT` directly in the environment rather
>>>> than in the config file.
>>>>
>>>> The reason I don't like (b) much is that we have to deal with the case
>>>> where somebody runs `path/foo/emcc` on the command line with
>>>> `EMSCRIPTEN_ROOT=path/bar` in the environment. We already have this issue
>>>> with `EMSCRIPTEN_ROOT` in the config file, and I'd love to remove it one
>>>> day.
>>>>
>>>> Regardless, unless I'm misunderstanding this is unrelated to this
>>>> thread which is about putting python literals in the EM_CONFIG
>>>> environment.. You aren't relying on that are you?
>>>>
>>>>
>>>>
>>>>
>>>>> Cheers!
>>>>> On Saturday, 10 April 2021 at 17:36:34 UTC+2 s...@google.com wrote:
>>>>>
>>>>>> FYI, I'm hoping to remove this feature this week:
>>>>>> https://github.com/emscripten-core/emscripten/pull/13855
>>>>>>
>>>>>> IIUC nobody is relying on it anymore.
>>>>>>
>>>>>> On Mon, Jul 20, 2020 at 8:57 AM Sam Clegg <s...@google.com> wrote:
>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On Thu, May 21, 2020 at 11:10 AM 'Steven Johnson' via
>>>>>>> emscripten-discuss <emscripte...@googlegroups.com> wrote:
>>>>>>>
>>>>>>>> It will be next week at the earliest -- I am currently elbow-deep
>>>>>>>> in work on the Halide build/test setup (and, as it turns out, the emcc
>>>>>>>> testing hasn't been enabled for our public repo recently, so it might
>>>>>>>> have
>>>>>>>> broken for other reasons).
>>>>>>>>
>>>>>>>>
>>>>>>> Hi Steven,
>>>>>>>
>>>>>>> I'm hoping to re-land this deprecation this week:
>>>>>>> https://github.com/emscripten-core/emscripten/pull/11684
>>>>>>>
>>>>>>> Have you managed to transition away from using this feature yet?
>>>>>>>
>>>>>>> cheers,
>>>>>>> sam
>>>>>>>
>>>>>>>
>>>>>>> On Thu, May 21, 2020 at 8:59 AM 'Sam Clegg' via emscripten-discuss <
>>>>>>>> emscripte...@googlegroups.com> wrote:
>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On Wed, May 20, 2020 at 6:55 PM 'Steven Johnson' via
>>>>>>>>> emscripten-discuss <emscripte...@googlegroups.com> wrote:
>>>>>>>>>
>>>>>>>>>> I did not know that. That sounds like a much better approach
>>>>>>>>>> anyway :-)
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> I'll hold off changing any of this until you can make your change
>>>>>>>>> downstream. Can you let me know when have converted?
>>>>>>>>>
>>>>>>>>> cheers,
>>>>>>>>> sam
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> On Wed, May 20, 2020 at 5:21 PM 'Sam Clegg' via
>>>>>>>>>> emscripten-discuss <emscripte...@googlegroups.com> wrote:
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On Wed, May 20, 2020 at 2:12 PM 'Steven Johnson' via
>>>>>>>>>>> emscripten-discuss <emscripte...@googlegroups.com> wrote:
>>>>>>>>>>>
>>>>>>>>>>>> Halide uses the EM_CONFIG variable in its build script.
>>>>>>>>>>>>
>>>>>>>>>>>> The reason for this is that we need to ensure that LLVM_ROOT
>>>>>>>>>>>> points to the specific version of LLVM being used by the rest of
>>>>>>>>>>>> Halide.
>>>>>>>>>>>> (Having these out of sync can cause amusingly bad things to
>>>>>>>>>>>> happen.) We
>>>>>>>>>>>> slurp in the default ~/.emscripten config file and make an altered
>>>>>>>>>>>> version
>>>>>>>>>>>> (with LLVM_ROOT pointing to the LLVM we are using), so that we can
>>>>>>>>>>>> always
>>>>>>>>>>>> build with the 'correct' version (ie the one that the rest of
>>>>>>>>>>>> Halide is
>>>>>>>>>>>> using).
>>>>>>>>>>>>
>>>>>>>>>>>> Requiring that ~/.emscripten be edited is problematic as Halide
>>>>>>>>>>>> supports multiple versions of LLVM at any one time (currently, v9,
>>>>>>>>>>>> v10, and
>>>>>>>>>>>> v11/trunk), so this file would need to edited in lockstep with
>>>>>>>>>>>> other things
>>>>>>>>>>>> in Halide to keep things in step.
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Thanks for the heads up steven!
>>>>>>>>>>>
>>>>>>>>>>> We had another report of someone using it, so I reverted the
>>>>>>>>>>> warning for now.
>>>>>>>>>>>
>>>>>>>>>>> I'm hoping we can move both of you in different direction over
>>>>>>>>>>> time. For example, did you know that you can set the LLVM_ROOT
>>>>>>>>>>> using the
>>>>>>>>>>> `EM_LLVM_ROOT` environment variable? This seems more targeted and
>>>>>>>>>>> easier
>>>>>>>>>>> for you. Any variable that you can set from the config file can
>>>>>>>>>>> also be
>>>>>>>>>>> set via the environment in the form of `EM_<NAME>`.
>>>>>>>>>>>
>>>>>>>>>>> cheers,
>>>>>>>>>>> sam
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> On Tue, May 19, 2020 at 11:46 AM 'Sam Clegg' via
>>>>>>>>>>>> emscripten-discuss <emscripte...@googlegroups.com> wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> emscripten has a feature where you can avoid using a config
>>>>>>>>>>>>> file
>>>>>>>>>>>>> completely and instead pass a python literal in the EM_CONFIG
>>>>>>>>>>>>> environment variable or `--em-config` command line.
>>>>>>>>>>>>>
>>>>>>>>>>>>> e.g. `emcc --em-config="<python code>`
>>>>>>>>>>>>>
>>>>>>>>>>>>> As part of routine cleanup I'm hoping to be able to remove this
>>>>>>>>>>>>> feature and allow the code to depend on the existence of a
>>>>>>>>>>>>> config
>>>>>>>>>>>>> file on disk.
>>>>>>>>>>>>>
>>>>>>>>>>>>> I'm not imagining there are any users of this feature, and we
>>>>>>>>>>>>> can
>>>>>>>>>>>>> make it into warning a few releases to confirm this.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Does anyone know of any users of this feature?
>>>>>>>>>>>>>
>>>>>>>>>>>>> cheers,
>>>>>>>>>>>>> sam
>>>>>>>>>>>>>
>>>>>>>>>>>>> --
>>>>>>>>>>>>> You received this message because you are subscribed to the
>>>>>>>>>>>>> Google Groups "emscripten-discuss" group.
>>>>>>>>>>>>> To unsubscribe from this group and stop receiving emails from
>>>>>>>>>>>>> it, send an email to emscripten-disc...@googlegroups.com.
>>>>>>>>>>>>> To view this discussion on the web visit
>>>>>>>>>>>>> https://groups.google.com/d/msgid/emscripten-discuss/CAL_va28seE50BNe4BKJ4cFzoQNNZPE6xbFFwDvXQSdWfiF2WAA%40mail.gmail.com
>>>>>>>>>>>>> <https://groups.google.com/d/msgid/emscripten-discuss/CAL_va28seE50BNe4BKJ4cFzoQNNZPE6xbFFwDvXQSdWfiF2WAA%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>>>>>>>>>>> .
>>>>>>>>>>>>>
>>>>>>>>>>>> --
>>>>>>>>>>>> You received this message because you are subscribed to the
>>>>>>>>>>>> Google Groups "emscripten-discuss" group.
>>>>>>>>>>>> To unsubscribe from this group and stop receiving emails from
>>>>>>>>>>>> it, send an email to emscripten-disc...@googlegroups.com.
>>>>>>>>>>>> To view this discussion on the web visit
>>>>>>>>>>>> https://groups.google.com/d/msgid/emscripten-discuss/CAM%3DdnveGTKtEqGWue_qyUMjdyLdGXZbXEm8m4fhPjreRgd4tGQ%40mail.gmail.com
>>>>>>>>>>>> <https://groups.google.com/d/msgid/emscripten-discuss/CAM%3DdnveGTKtEqGWue_qyUMjdyLdGXZbXEm8m4fhPjreRgd4tGQ%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>>>>>>>>>> .
>>>>>>>>>>>>
>>>>>>>>>>> --
>>>>>>>>>>> You received this message because you are subscribed to the
>>>>>>>>>>> Google Groups "emscripten-discuss" group.
>>>>>>>>>>> To unsubscribe from this group and stop receiving emails from
>>>>>>>>>>> it, send an email to emscripten-disc...@googlegroups.com.
>>>>>>>>>>> To view this discussion on the web visit
>>>>>>>>>>> https://groups.google.com/d/msgid/emscripten-discuss/CAL_va29xtsd5emac0nfmvUys84xaBdx-xMes515U1sATFPJe1w%40mail.gmail.com
>>>>>>>>>>> <https://groups.google.com/d/msgid/emscripten-discuss/CAL_va29xtsd5emac0nfmvUys84xaBdx-xMes515U1sATFPJe1w%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>>>>>>>>> .
>>>>>>>>>>>
>>>>>>>>>> --
>>>>>>>>>> You received this message because you are subscribed to the
>>>>>>>>>> Google Groups "emscripten-discuss" group.
>>>>>>>>>> To unsubscribe from this group and stop receiving emails from it,
>>>>>>>>>> send an email to emscripten-disc...@googlegroups.com.
>>>>>>>>>> To view this discussion on the web visit
>>>>>>>>>> https://groups.google.com/d/msgid/emscripten-discuss/CAM%3Ddnvf79x%3DXt_CLym76VCoTeJBmpu0pw9%2BWWR0CH6ixTcjptg%40mail.gmail.com
>>>>>>>>>> <https://groups.google.com/d/msgid/emscripten-discuss/CAM%3Ddnvf79x%3DXt_CLym76VCoTeJBmpu0pw9%2BWWR0CH6ixTcjptg%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>>>>>>>> .
>>>>>>>>>>
>>>>>>>>> --
>>>>>>>>> You received this message because you are subscribed to the Google
>>>>>>>>> Groups "emscripten-discuss" group.
>>>>>>>>> To unsubscribe from this group and stop receiving emails from it,
>>>>>>>>> send an email to emscripten-disc...@googlegroups.com.
>>>>>>>>> To view this discussion on the web visit
>>>>>>>>> https://groups.google.com/d/msgid/emscripten-discuss/CAL_va2_CFGdXYRhbB2Vx3Cim8qo0Z1NuErepc-GapsAmDWNBLQ%40mail.gmail.com
>>>>>>>>> <https://groups.google.com/d/msgid/emscripten-discuss/CAL_va2_CFGdXYRhbB2Vx3Cim8qo0Z1NuErepc-GapsAmDWNBLQ%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>>>>>>> .
>>>>>>>>>
>>>>>>>> --
>>>>>>>> You received this message because you are subscribed to the Google
>>>>>>>> Groups "emscripten-discuss" group.
>>>>>>>> To unsubscribe from this group and stop receiving emails from it,
>>>>>>>> send an email to emscripten-disc...@googlegroups.com.
>>>>>>>> To view this discussion on the web visit
>>>>>>>> https://groups.google.com/d/msgid/emscripten-discuss/CAM%3DdnvcWFdmoiG-1-z2yhPqgEHdSXs5aRHJpc%2BTtQXL26Z0pGA%40mail.gmail.com
>>>>>>>> <https://groups.google.com/d/msgid/emscripten-discuss/CAM%3DdnvcWFdmoiG-1-z2yhPqgEHdSXs5aRHJpc%2BTtQXL26Z0pGA%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>>>>>> .
>>>>>>>>
>>>>>>> --
>>>>> You received this message because you are subscribed to the Google
>>>>> Groups "emscripten-discuss" group.
>>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>>> an email to emscripten-disc...@googlegroups.com.
>>>>>
>>>> To view this discussion on the web visit
>>>>> https://groups.google.com/d/msgid/emscripten-discuss/185eae5b-2ee6-462d-8719-1210204d4f5cn%40googlegroups.com
>>>>> <https://groups.google.com/d/msgid/emscripten-discuss/185eae5b-2ee6-462d-8719-1210204d4f5cn%40googlegroups.com?utm_medium=email&utm_source=footer>
>>>>> .
>>>>>
>>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "emscripten-discuss" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to emscripten-disc...@googlegroups.com.
>>>
>> To view this discussion on the web visit
>>> https://groups.google.com/d/msgid/emscripten-discuss/2653d96f-9437-449b-a0dd-7938d0fdd7e9n%40googlegroups.com
>>> <https://groups.google.com/d/msgid/emscripten-discuss/2653d96f-9437-449b-a0dd-7938d0fdd7e9n%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>> --
> You received this message because you are subscribed to the Google Groups
> "emscripten-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to emscripten-discuss+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/emscripten-discuss/6be4ea24-59cd-4a96-9c75-644a1c18e395n%40googlegroups.com
> <https://groups.google.com/d/msgid/emscripten-discuss/6be4ea24-59cd-4a96-9c75-644a1c18e395n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
--
You received this message because you are subscribed to the Google Groups
"emscripten-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to emscripten-discuss+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/emscripten-discuss/CAL_va2_0q7_bbWachjbpwu1qrDzdAiJbYwtpu%2BQxZ7%3D6uK5SCw%40mail.gmail.com.
