In a change committed yesterday, the v8 shell was disabled with the comment "do not install shell (it's not intended to be installed; d8 is)".
I tried creating a patch to add a "d8" use flag, but it didn't work terribly well, and kept installing d8 to the root of the image directory. I've attached what I got in case it's helpful, and since I'd like to at least get some guidance on what I'm doing wrong so I can learn from the experience.
Steps to Reproduce:
1. emerge =dev-lang/v8-22.214.171.124
d8: command not found
V8 version 126.96.36.199 [D8 light using shared library]
Created attachment 289781 [details, diff]
attempt to patch in a d8 use flag
A more generic option for the use flag might be something like 'shell' instead of 'd8'. I'm hoping to have a similar one added to dev-lang/spidermonkey to re-institute the 'js' shell.
v8-3.4 doesn't even build d8 because it's broken upstream. v8-3.6 (hard masked) builds and installs v8. Sorry for the inconvenience, feel free to tell upstream what you think.
By the way, v8-3.4 is *not* affected by the change you've mentioned (really, please see the ChangeLog and the list of touched files above that message).
Ah, the changelog is definitely my mistake. Sorry about that.
I am still slightly confused about why when I do a "scons d8" the resulting executable seems to work (at least runs - I understand that's not even remotely close to the same thing). I will take my search upstream. Sorry if I seemed at all accusatory - I really appreciate all the hard work you're all putting in here, and will definitely be looking at 4.6.
(In reply to comment #4)
> I am still slightly confused about why when I do a "scons d8" the resulting
> executable seems to work (at least runs - I understand that's not even remotely
> close to the same thing).
A possible explanation: Simply running "scons d8" results in a statically linked version of v8, whereas the ebuild builds a dynamically linked (shared) version. Upstream tends to take better care of the static version since it is used by the Chrome web browser.
Ah, I see. Thanks for the reply, and makes sense.
Also, pleased to report that upgrading to the 3.6* line did definitely bring back a working d8 binary, in case anyone stumbles across this ticket and doesn't catch that. Again, thanks for the good work guys.