Bug 1717198 Comment 2 Edit History

Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.

The chances of these regressions really having been caused by the NodeJS upgrade strike me as pretty remote, 

Unfortunately, I'm out on PTO, so I'm going to have to depend on some folks here.

First off, my sense is that backing this out should be avoided at all possible, because we've just asked all the developers to upgrade their toolchains to Node 12, and since --enable-bootstrap was landed after this patch, if we back it out, everybody will have to downgrade by running `mach bootstrap` manually, which sounds exceedingly painful.  mhentges, would you agree?  If so, can you help prevent that from happening??

The main thing that the NodeJS upgrade would effect in terms of what gets built on the build machines, as far as I know, is the bundling of devtools.  That strikes me as quite unlikely to have caused the perf regressions linked to here, but maybe there's something I don't know about how the devtools code works.  Honza, what are your any thoughts on this?

One thing that springs to mind here is if PerfHerder itself uses the in-tree Node toolchain (eg a Node script does the testing and timing using web driver or puppeteer or something), and somehow changing versions of Node changed that timing.  The chances of this strike me as fairly low, but I've been wrong before!  :florin, would you please reach out to a perfherder owner to see if that theory is even slightly plausible, or if they have any better suggestions of things to look into?

Mardak, Standard8, can you guys think of anything that could explain this?

mhentges, would you be willing to take this bug in my absence and drive it?

Thanks to all for any and all help/input..
FWIW, the chances of these regressions really having been caused by the NodeJS upgrade strike me as pretty low.

Unfortunately, I'm out on PTO, so I'm going to have to depend on some folks here.

First off, my sense is that backing this out should be avoided at all possible, because we've just asked all the developers to upgrade their toolchains to Node 12, and since --enable-bootstrap was landed after this patch, if we back it out, everybody will have to downgrade by running `mach bootstrap` manually, which sounds exceedingly painful.  mhentges, would you agree?  If so, can you help prevent that from happening?

The main thing that the NodeJS upgrade would effect in terms of what gets built on the build machines, as far as I know, is the bundling of devtools. That strikes me as quite unlikely to have caused the perf regressions linked to here, but maybe there's something I don't know about how the devtools code works. Honza, what are your thoughts on this?

One possibility that springs to mind is if PerfHerder itself uses the in-tree Node toolchain (eg a Node script does the testing and timing using web driver or puppeteer or something), and somehow changing versions of Node changed that timing. The chances of this strike me as fairly low, but I've been wrong before!  :Bebe, would you please reach out to a perfherder owner and ask them if that theory is even slightly plausible, or if they have any better suggestions of things to look into?

Mardak, Standard8, can you guys think of anything that could explain this?

mhentges, would you be willing to take this bug in my absence and drive it?

Thanks to all for any and all help/input..

Back to Bug 1717198 Comment 2