915129 - Don't link in epoll_sub.c from libevent on Android or B2G.

Status: RESOLVED FIXED

(Core :: IPC, defect)

Description

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

Attachment: rm-epoll-sub.diff

Android's libc has syscall stubs for epoll_{create,ctl,wait}, so we don't need libevent's replacements that call syscall().  More importantly, Android's libc has unwinding info for epoll_* but not syscall(), so this will be a very noticeable papercut when bug 810526 lands.

But we do need them when building for regular Linux, because otherwise it breaks the build, because apparently we support a relatively old glibc.  (Symbol versions suggest that 2.3.2 added them.)  libevent's configure script would deal with this, but we're not running it.  Thus, this patch.

https://tbpl.mozilla.org/?tree=Try&rev=651fee590a65

See also bug 905521 for discussion of the bionic side of this.

Comment 1

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Comment on attachment 802963 [details] [diff] [review]
rm-epoll-sub.diff

rs=me if this builds and passes tests :)

Comment 2

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

https://tbpl.mozilla.org/?tree=Try&rev=9012e2011db1 has passed everything except the Android 2.2 tests which look like they have a huge backlog (and haven't even started yet), and if this change was going to fail at all it was almost certainly going to break the build, and in fact the builds all passed.  So I'm feeling confident about the cost/benefit here.

For checkin to b2g-inbound.

Comment 3

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/96e921ddc1fd

Comment 4

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/96e921ddc1fd