Should we have a setting that is the opposite of `LEGACY_VM_SUPPORT`

[sbc100]

i.e. a way to say "I want to target only the current latest browsers".

Maybe there is some way to do this today, but I don't know what it is.

Setting -sMIN_CHROME_VERSION=-1, etc, doesn't quite work because that says we don't want support chrome at all.

Perhaps the explicit -sLEGACY_VM_SUPPORT=0 could set all of the MIN versions to some huge value? Or is that too magical?

Perhaps a new setting list -sBLEEDING_EDGE_VM_SUPPORT_ONLY.. i can't think if a good/short name for that though.

[sbc100]

The main reason I want such as setting is for testings, so its easy to see what the emscripten output looks like when it not trying to work around old browser issues.

[sbc100]

@kripken @dschuff @juj

[kripken]

-sMODERN_VM_SUPPORT? Seems reasonable to me as a setting.

[sbc100]

I think the name should indicate that we drop support for all but the latest browsers.

Emscripten always supports modern browsers, so this is more about dropping support for older browser. -sMODERN_VM_ONLY?

-sLATEST_VM_ONLY?

[dschuff]

Just to be clear, would this mean:

1. Support the latest stable version of all browsers
2. Support the latest stable version of the browser with the most features (Chromium usually)
3. Enable all the features (which could be a superset of any one browser, or include features not released yet)

[sbc100]

I think I was thinking (1).

Or more precisely, for any browser that is not explicitly specified assume the latest stable version. Basically push all the defaults to "latest".

That way if somebody wants to do (3) then they could just disable FF & safari.

One issue is there is no current magic "latest" value I can use MIN_XX_VERSION like there is for "unsupported" (-1). I guess I'm proposing to add something like `-sMIN_XX_VERSION=latest"

[dschuff]

-sMIN_CHROME_VERSION=∞ :D

[dschuff]

As a practical matter if we do something like this, I think it should be pinned (e.g. in the source) rather than dynamically updating based on what is the latest browser at the time of compilation. e.g. Emscripten version 6.0.x should always use the latest version as of the time of its release, because having the behavior of the compiler under a constant set of flags change over time seems bad.

[sbc100]

Yes, the idea is basically, "assume all features that we are aware of exist". Obviously we cannot support new feature we are not aware of.

[dschuff]

but "all features we are aware of" (i.e. my (3) above) is very different from (1). Emscripten will be aware of features for a long time before they are supported in all browsers.

[sbc100]

Sorry, yes you are correct. Of a given feature is simply not support by browser X, and you don't disable browser X then this flag will not opt into it.

Basically any line like would be come true under this new flag:

#if MIN_CHROME_VERSION > 123 && MIN_FIRE_FIRFOX_VERSION > 345

But this line would not become true (unless you also opt-out of firefox support).

#if MIN_CHROME_VERSION > 123 && MIN_FIRE_FIRFOX_VERSION != UNSUPPORTED

I'm not proposing that we imagine browsers supporting features they do not.

One way to implement this feature would be:

LATEST_VERSION = 0xffffffff
default_setting('MIN_CHROME_VERSION', LATEST_VERSION)
default_setting('MIN_FIREFOX_VERSION', LATEST_VERSION)
default_setting('MIN_SAFARI_VERSION', LATEST_VERSION)

Or alternatively something like:

default_setting('MIN_CHROME_VERSION', MAX_KNOWN_CHROME_VERSION)
default_setting('MIN_FIREFOX_VERSION', MAX_KNOWN_FIREFOX_VERSION)
default_setting('MIN_SAFARI_VERSION', MAX_KNOWN_SAFARI_VERSION)

I think the effect would be the same either way.