Handle multipart/form-data submissions on native crashes in Linux

[implausible]

Fixes #200. I think what was not mentioned in the issue is that currently Sentry does not handle these submissions at all, and so native crash reports on Linux are currently being labeled as corrupted submissions in Sentry. We need this solved ASAP so that we can adopt the product.

This adds the ability to detect whether a crash report is a multipart/form-data submission body and to switch minidump upload behavior to accommodate that on Linux. I have the multipart/form-data being sent to the /api/{projectId}/minidump route instead of the /api/{projectId}/envelope route, as when I tried uploading the minidump with the sentry event attached to it, sentry's servers returned a 400 status code, causing the system to throw. If it was intended to allow the envelope route to accept the multipart/form-data submission, it might need to be fixed server side to return a 200 OK in response.

I have adjusted the multipart/form-data submission to include the event payload under the name sentry as suggested in the issue. From testing against sentry, I can confirm that this produces the desired behavior where the full sentry event and minidump attachment are properly received on sentry.

Due to the changes introduced here, I went ahead and adjusted the test server to accept minidump multipart/form-data submissions. I also reenabled all tests as well as addressed the blocking issues from running tests on 10 and up. I was able to reproduce the issues that caused the team to disable specs on 9.0.5. I traced the failure to this issue electron/electron#24788. This was addressed in 9.2.0. I would advise that for anyone affected by that issue, the team suggests to the affected user to move their Electron target.

Please let me know if there is anything I can do to speed up the process of getting this merged / released. My team is looking to adopt Sentry for crash reporting and need this solved.

[implausible]

I've disabled testing on Electron < 4 due to some stability issues that exist in the windows e2e tests. While it's not really my decision what your minimum support is, Electron has a much smaller support target than it maintained in the past, with versions < 9 no longer being supported by the team.

[timfish]

I was doing some Electron v11 testing locally today and saw the same test onFatalError test failure on macOS too.

[implausible]

I was doing some Electron v11 testing locally today and saw the same test onFatalError test failure on macOS too.

I am not sure that this is a fault of this branch though. From an improvement perspective, I believe this branch covers a condition that the sentry electron sdk currently does not support without introducing regressions. Are there any changes you would like me to make? I'd love to see this merged and released, so that I can stop relying on a fork for sentry support in Electron 10.

[timfish]

The preload tests are failing due to this erroneous line which I think was left on master by mistake. I've removed it in my recent un-merged PRs and it fixes the failures.

[timfish]

@implausible if you can rebase this to master I can get it merged!

[implausible]

I will get to this today

[timfish]

I managed to do the merge via GitHub web UI. Just looking into tests...

[implausible]

I managed to do the merge via GitHub web UI. Just looking into tests...

Oh, awesome!

[timfish]

@implausible it looks like there are many failing tests for minidumps.

Some debugging shows that the envelope body is not parsed correctly by the test server and this looks to be caused by the changes to net.ts passing the content type.

[timfish]

Closing in favour of #343 but thank you for your efforts!