366113 - mozilla-plugin.pc should not depend on mozilla-xpcom.pc

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[Braden]

Attachment: Patch

mozilla-plugin.pc's dependency on mozilla-xpcom.pc is obsolete. Persons who need to compile plug-ins that need XPCOM can use mozilla-xpcom.pc directly; meanwhile, plug-ins that don't need XPCOM shouldn't be linking with libxpcom.

Comment 1

[Benjamin Smedberg]

Comment on attachment 250676 [details] [diff] [review]
Patch

By the same plugin, why would these plugins require NSPR?

Comment 2

[Benjamin Smedberg]

by the same logic, even

Comment 3

[Braden]

npapi.h includes prtypes.h.

Comment 4

[Benjamin Smedberg]

Comment on attachment 250676 [details] [diff] [review]
Patch

That's... unfortunate. Can that be fixed?

Comment 5

[Braden]

I think so--as long as we don't care about source compatibility with stuff that compiles against npapi.h today.

Bug 366858.

Comment 6

[Nickolay_Ponomarev]

checked in:
mozilla/build/unix/mozilla-plugin.pc.in  1.3

Comment 7

[Daniel Veditz [:dveditz]]

Why should we be taking this on the branches? What's the impact, risk, gain?

Comment 8

[Braden]

Risk is negligible, if it exists at all. In the worst case, build systems that were relying on mozilla-plugin.pc to pull in libxpcom build flags will have to be modified to do that a different way (presumably by using mozilla-xpcom.pc explicitly).

The gain is that clients using mozilla-plugin.pc that do *not* use XPCOM will no longer incur a spurious dependency on libxpcom.

Comment 9

[Daniel Veditz [:dveditz]]

Comment on attachment 250676 [details] [diff] [review]
Patch

approved for 1.8/1.8.0 branches, a=dveditz for drivers

Comment 10

[Jay Patel [:jay]]

Braden:  We need this landed asap if it's going to make the release, so if you can find someone to land this for you before Monday morning, that'll great!  Maybe bsmedberg?

Comment 11

[Braden]

This has been checked in on the 1.8 and 1.8.0 branches.

Comment 12

[Christopher Aillon (sabbatical, not receiving bugmail)]

For the record, this is the type of stuff that should NOT be landing on branches after there has been a release and people are dependent on the behavior.  Several customers have already complained.  The fact that this was not even release noted is just plain sloppy.