612439 - Linux-deep compile failure in nanojit

Status: RESOLVED FIXED

(Tamarin Graveyard :: Baseline JIT (CodegenLIR), defect, P3)

Description

[James Sudduth]

There is a compile failure in linux-deep:

/home/build/buildbot/tamarin-redux/linux-deep/repo/core/AvmCore.cpp: In static member function ‘static int avmplus::AvmCore::integer_d(double)’:
/home/build/buildbot/tamarin-redux/linux-deep/repo/core/AvmCore.cpp:4179: error: comparison between signed and unsigned integer expressions
/home/build/buildbot/tamarin-redux/linux-deep/repo/core/../nanojit/Nativei386.h: At global scope:
/home/build/buildbot/tamarin-redux/linux-deep/repo/core/../nanojit/Nativei386.h:167: error: ‘nanojit::ScratchRegs’ defined but not used
/home/build/buildbot/tamarin-redux/linux-deep/repo/core/../nanojit/Nativei386.h:170: error: ‘nanojit::AllowableByteRegs’ defined but not used

There are more logs here;

http://tamarin-builds.mozilla.org/tamarin-redux/builders/linux-deep/builds/19

Comment 1

[Edwin Smith]

is this because we're using different warnings or a different version of GCC?  even the main linux build could catch these, maybe.

AvmCore.cpp:4179 is

  int intval = ...
  if (intval != 0x8000000)

so a simple cast would easily fix it.  hg annotate points to this changeset:

changeset:   5494:c9c586e0b08d
user:        Werner Sharp ([e-mail])
date:        Thu Nov 04 12:13:53 2010 -0700
summary:     bug 608801.  remove real2int.  (r=stejohns)

Comment 2

[Brent Baker]

(In reply to comment #1)
> is this because we're using different warnings or a different version of GCC? 
> even the main linux build could catch these, maybe.

Wonder if this is a new warning in 4.3.3? The linux32 main build is using 4.2.1 and the main linux64 build is using 4.3.2

Comment 3

[Edwin Smith]

Possibly.  These should be trivial fixes - the best thing to do is for the person with access to the target machine to create a patch and post for review.

As followup, can the main build to be updated to the newest/strictest GCC, so we catch trivial problems sooner?

Comment 4

[James Sudduth]

Assigning to myself to drive closure.

Comment 5

[James Sudduth]

Turns out the AvmCore.cpp part is easy to fix but the Nativei386.h part will require effort from someone familiar with nanojit. Reassigning to Rick.

As for the question in comment 3 about updating the main builders please see bug 614059.

Comment 6

[Trevor Baker]

Rick, there is talk of a merge to mainline this week.  I assume complication failures are blockers...

Comment 7

[Rick Reitmaier]

What is blocking?  I see a failure with 'no-jit' configuration.  As none of our products use this config does this really block anything?   

I agree it would be nice to clean up eventually.

Comment 8

[Trevor Baker]

There are two other failing steps, compiling and testing:

ReleaseDebugger-air
Release-system-malloc

All failing in same fashion.

Comment 9

[Brent Baker]

Attachment: add a cast to fix compiler warning

Comment 10

[Trevor Baker]

Comment from Rick via email:
>>>It looks like this requires not only fixes in tamarin but also a nanojit merge meaning that its most likely not going to be fixed and merged by tomorrow.
Why is a high priority, it looks like these are builds that exclude the JIT?

Comment 11

[Rick Reitmaier]

Comment on attachment 496132 [details] [diff] [review]
add a cast to fix compiler warning

r+ should be int32_t cast not int.

Comment 12

[Rick Reitmaier]

(In reply to comment #11)
> Comment on attachment 496132 [details] [diff] [review]
> add a cast to fix compiler warning
> 
> r+ should be int32_t cast not int.

I take back the nit, the cast should be int, to match the type of intval.

Comment 13

[Tamarin Bot]

changeset: 5645:e2fdf6655530
user:      Brent Baker <[e-mail]>
summary:   Bug 612439: fix compilation issues with a cast (r=rreitmai)

http://hg.mozilla.org/tamarin-redux/rev/e2fdf6655530