400759 - PSM should build with NSS size optimization

Status: RESOLVED FIXED

(Core :: Security: PSM, defect)

Description

[Kai Engert [:KaiE:]]

PSM should build with NSS size optimization

Spin off from bug 389781.

Comment 1

[Kai Engert [:KaiE:]]

Attachment: Patch v1

Comment 2

[Robert Relyea]

Comment on attachment 285789 [details] [diff] [review]
Patch v1

r+ rrelyea

Comment 3

[Wan-Teh Chang]

Comment on attachment 285789 [details] [diff] [review]
Patch v1

r=wtc.

Comment 4

[Wan-Teh Chang]

The patch for bug 389781 only honors the OPT_CODE_SIZE flag on
Mac, Linux, and Windows.  It would be nice for Mozilla developers
to extend that patch to other platforms.  In particular it should
be easy to extend it to the *BSDs because they all use GCC.

Comment 5

[Mike Beltzner [:beltzner, not reading bugmail]]

Comment on attachment 285789 [details] [diff] [review]
Patch v1

a=endgame drivers for M9

Comment 6

[Kai Engert [:KaiE:]]

landed for M9, marking fixed.

Comment 7

[Boris Zbarsky [:bzbarsky]]

This checkin had absolutely no effect on the Tinderbox codesize numbers.  Then again, I also don't see the NSS things I'd expect to rebuild after this checkin being rebuilt.

Reopening pending that getting sorted out (e.g. with a clobber if that's what's needed).

Comment 8

[Reed Loden [:reed]]

Information from Linux clobber:

BEFORE:
Z:14.31MB
mZ:2.719MB

AFTER:
Z:13.94MB
Zdiff:-389954 (+33327/-423281)
mZ:2.372MB
mZdiff:-363554 (+32822/-396376)

Comment 9

[Reed Loden [:reed]]

Information from Mac clobber:

BEFORE:
Z:14.92MB

AFTER:
Z:14.67MB
Zdiff:-262144 (+68562/-330706)

Comment 10

[Boris Zbarsky [:bzbarsky]]

OK, re-resolving.

Comment 11

[Nelson Bolyard (seldom reads bugmail)]

Reed, please provide an interpretation of the results you posted.
Z?  mZ?  
Zdiff?  mZdiff?  (are those code size differences?)

Comment 12

[Reed Loden [:reed]]

Z -- Firefox: Code + data size of all shared libs & executables
mZ -- Embed: Code + data size of all shared libs & executables

The Zdiff and mZdiff display any differences between the last values of Z and mZ and their current values.

Comment 13

[Boris Zbarsky [:bzbarsky]]

For Firefox, the mZ number is irrelevant.  The Z number is the only number that makes sense.

The Z number is the sum of the sizes of the code and data segments (according to readelf, iirc) in all the executables and shared libraries that we actually ship.