Prevent packing of native modules

[PF4Public]

Fixes element-hq/element-web#17188

before:

 * Searching for element-desktop ...
 * Contents of net-im/element-desktop-9999:
/usr
/usr/lib64
/usr/lib64/element-desktop
/usr/lib64/element-desktop/app.asar
/usr/lib64/element-desktop/img
/usr/lib64/element-desktop/img/element.ico
/usr/lib64/element-desktop/img/element.png
/usr/lib64/element-desktop/webapp -> ../../share/element-web
/usr/share
/usr/share/applications
/usr/share/applications/electron-16-element-desktop.desktop

after:

 * Searching for element-desktop ...
 * Contents of net-im/element-desktop-1.10.9:
/usr
/usr/lib64
/usr/lib64/element-desktop
/usr/lib64/element-desktop/app.asar
/usr/lib64/element-desktop/app.asar.unpacked
/usr/lib64/element-desktop/app.asar.unpacked/node_modules
/usr/lib64/element-desktop/app.asar.unpacked/node_modules/keytar
/usr/lib64/element-desktop/app.asar.unpacked/node_modules/keytar/build
/usr/lib64/element-desktop/app.asar.unpacked/node_modules/keytar/build/Release
/usr/lib64/element-desktop/app.asar.unpacked/node_modules/keytar/build/Release/keytar.node
/usr/lib64/element-desktop/app.asar.unpacked/node_modules/matrix-seshat
/usr/lib64/element-desktop/app.asar.unpacked/node_modules/matrix-seshat/native
/usr/lib64/element-desktop/app.asar.unpacked/node_modules/matrix-seshat/native/index.node
/usr/lib64/element-desktop/img
/usr/lib64/element-desktop/img/element.ico
/usr/lib64/element-desktop/img/element.png
/usr/lib64/element-desktop/webapp -> ../../share/element-web
/usr/share
/usr/share/applications
/usr/share/applications/electron-18-element-desktop.desktop

---

Here's what your changelog entry will look like:

🐛 Bug Fixes

• Prevent packing of native modules (#337). Fixes element-desktop creates /tmp/.org.chromium.Chromium.xxxxxx files when started, which are not deleted upon exit element-web#17188. Contributed by @PF4Public.

[ArchangeGabriel]

Admittedly I was surprised that they managed to pack the native modules into the asar. So not surprised this might not have been a good idea, though I don’t remember why they were packed in the first place.

[PF4Public]

I'd guess it is not that natives were intentionally packaged, rather noone bothered not to do so.

[t3chguy]

This seems to double up the native modules as now they're both in and out of the asar.

[PF4Public]

@t3chguy Good catch! I'll look into it.

[PF4Public]

@t3chguy Have you tried extracting the resulting asar? I suppose it could be so "by design".

I've done some testing and run this after each rebuild:

node node_modules/.bin/asar l dist/linux-unpacked/resources/app.asar | grep \\.node$;
ls -lh dist/linux-unpacked/resources;
node node_modules/.bin/asar e dist/linux-unpacked/resources/app.asar extrrrrrr;

First without my modifications the result was as follows:

/node_modules/matrix-seshat/native/index.node
/node_modules/keytar/build/Release/keytar.node
total 32M
-rw-r--r-- 1 portage portage  32M Apr 13 22:14 app.asar
drwxr-xr-x 2 portage portage 4.0K Apr 13 22:14 img

After repeating with asarUnpack I got this:

/node_modules/matrix-seshat/native/index.node
/node_modules/keytar/build/Release/keytar.node
total 25M
-rw-r--r-- 1 portage portage  25M Apr 13 22:19 app.asar
drwxr-xr-x 3 portage portage 4.0K Apr 13 22:19 app.asar.unpacked
drwxr-xr-x 2 portage portage 4.0K Apr 13 22:19 img

It is obviously lighter now. Natives could still be referenced in asar, but absent from it. Lets check this assumption by renaming app.asar.unpacked:

node node_modules/.bin/asar l dist/linux-unpacked/resources/app.asar | grep \\.node$;
mv dist/linux-unpacked/resources/app.asar.unpacked dist/linux-unpacked/resources/dhdfghsetrhsrthjdrthdrtj
ls -lh dist/linux-unpacked/resources;
node node_modules/.bin/asar e dist/linux-unpacked/resources/app.asar extrrrrrr;

It fails with a very descriptive error message :)

/node_modules/matrix-seshat/native/index.node
/node_modules/keytar/build/Release/keytar.node
total 25M
-rw-r--r-- 1 portage portage  25M Apr 13 22:22 app.asar
drwxr-xr-x 3 portage portage 4.0K Apr 13 22:22 dhdfghsetrhsrthjdrthdrtj
drwxr-xr-x 2 portage portage 4.0K Apr 13 22:22 img
internal/fs/utils.js:314
    throw err;
    ^

Error: ENOENT: no such file or directory, open '…snip…/dist/linux-unpacked/resources/app.asar.unpacked/node_modules/matrix-seshat/native/index.node'

Another interesting thing. I found this in electron-builder docs:

Node modules, that must be unpacked, will be detected automatically, you don’t need to explicitly set asarUnpack - please file an issue if this doesn’t work.

This suggests it could be also a bug in electron-builder 🤷

[t3chguy]

Thanks for the deep dive, can confirm this works and is a great one liner! Thanks

[t3chguy]

Needs dco sign off as per https://github.com/matrix-org/matrix-js-sdk/blob/develop/CONTRIBUTING.md#sign-off

[PF4Public]

If "legally identifiable name" includes "pseudonym", I would be happy to rewrite the commit and add Signed-off-by:.

[turt2live]

If it's legal where you're at, then yes.

[ArchangeGabriel]

Well given the point is that it cannot be anonymous, I don’t think that would work. What should be done in this case? Just someone else pushing the same change?

[turt2live]

@ArchangeGabriel we've accepted contributions from pseudonyms before, but are aware of several countries where it wouldn't be legally binding.

[PF4Public]

@t3chguy @turt2live Done.

Well given the point is that it cannot be anonymous, I don’t think that would work.

Pseudonymous is not anonymous.

What should be done in this case? Just someone else pushing the same change?

Probbaly this:

(c) The contribution was provided directly to me by some other
person who certified (a), (b) or (c) and I have not modified
it.

[robintown]

@PF4Public If you don't want to disclose your legal name publicly, another option is to email dco@matrix.org, who have a process in place for this kind of thing.

[PF4Public]

@PF4Public If you don't want to disclose your legal name publicly, another option is to email dco@matrix.org, who have a process in place for this kind of thing.

Thanks for suggestion! But I didn't contribute any significant mount of code, so I assume pseudonym suffice.

[ArchangeGabriel]

So we discussed that OOB, and pseudonymous is fine enough to not be anonymous here (I said that this would be really context dependent, because for all that matters a pseudonym could be completely untraceable to a physical person), especially given the small amount of change as stated.

[PF4Public]

pseudonym could be completely untraceable to a physical person

If I were to sign as follows: Signed-off-by: Nashorn Unpaarhuferson <NaUR@Upplýsingalæsi.Skuespiller.示例域> how am i to be traced?

[turt2live]

we're having this conversation elsewhere - please let's keep it off the PR.

We don't technically require signoff on this change, but it's borderline enough where it's warranted to ask for it. If we have any reason to believe this change was made in bad faith (stolen code, malicious intent, etc) then we'd handle it accordingly.

[PF4Public]

Whatever you decide. Feel free to submit this one-liner in the name of one of your developers. No problem. Just fix this year-long nuisance already :)

[t3chguy]