Thank you, Jakob - I've relayed those items of guidance to the Debian and GitHub threads, and I think that probably completes my involvement in the discussions.
On Fri, 10 Mar 2023 at 20:51, Jakob Kummerow <[email protected]> wrote: > > It seems that on 32-bit ARM, the original reason for the lower limit has > disappeared over time, and raising the stack limit there again to the same > 984KB as on ia32/x64 might be worth a shot. > > However, what Node/Debian cares about is arm64, and I see no reason to assume > that anything has changed there since the limit was lowered. So I wouldn't be > comfortable raising it. > > Since (per crbug.com/v8/10575) the issue was specific to WebView, maybe the > best path forward would be for Node to float a patch. V8 itself doesn't know > what product it's getting embedded into. > > FWIW, there is no guarantee that the same stack size allows the same > recursion depth on different platforms. In particular, 32-bit platforms will > allow significantly higher recursion depths than 64-bit platforms per > kilobyte of stack space, because most spilled values are register-sized. > There could also be platform-specific differences between different 64-bit > architectures, in particular in optimized code. > > The "proper" fix is to design JS applications such that they don't even > approach the stack limit. If 984 vs 864 makes a given app work or fail, then > it's walking on very thin ice either way. > > > On Fri, Mar 10, 2023 at 4:00 PM 'James Addison' via v8-dev > <[email protected]> wrote: >> >> Hi folks, >> >> Debian bug #1030284[1] and the related NodeJS GitHub issue #41163[2] >> report ARM-specific RangeError exceptions from the vendored V8 library >> in NodeJS. The bug(s) are reproducible with v18.13 of NodeJS. >> >> The cause of the difference-in-behaviour appears to be that V8 sets[3] >> a lower stack size for ARM: 864K as compared to 984K. >> >> Would it be safe to increase the stack size on ARM to 984K to restore >> consistency with most other architectures? >> >> (I've offered a patch to make that change, and others have confirmed >> that it allows a provided repro case to pass; I'm worried about any >> potential unsafe side-effects of the change, though) >> >> Thank you, >> James >> >> [1] - https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1030284 >> >> [2] - https://github.com/nodejs/node/issues/41163 >> >> [3] - >> https://github.com/nodejs/node/blob/2bb4b59fa5529569ad38d3bf7d36666c926d8e47/deps/v8/src/common/globals.h#L74-L86 >> >> -- >> -- >> v8-dev mailing list >> [email protected] >> http://groups.google.com/group/v8-dev >> --- >> You received this message because you are subscribed to the Google Groups >> "v8-dev" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To view this discussion on the web visit >> https://groups.google.com/d/msgid/v8-dev/CALDQ5NzeykWdpd79RcwDwJiWN-tYm%3DE1soUi8bxB1X%3DnjM5Z1g%40mail.gmail.com. > > -- > -- > v8-dev mailing list > [email protected] > http://groups.google.com/group/v8-dev > --- > You received this message because you are subscribed to the Google Groups > "v8-dev" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/v8-dev/CAKSzg3Tnw06NZJhSoE7YzJww0rX5%3DZU%3DYb63Gocuy%3Dz-49hw0Q%40mail.gmail.com. -- -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev --- You received this message because you are subscribed to the Google Groups "v8-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/v8-dev/CALDQ5NyU58FFaf46sjrF4-v7D2C1f-5qF2%2BXOU8hpSvgBiscEA%40mail.gmail.com.
