525878 - The lirasm "call1" test fails on ARM.

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jacob Bramley [:jbramley]]

Attachment: Fix lirasm.cpp so that it evaluates rint() and rfloat() before it starts to print the result.

==== I'm using lirasm from nanojit-central. ====

The lirasm.cpp test utility prints the output like this:

cout << "Output is: " << i->second.rint() << endl;

The rint() method appears to actually calculate the result. On x86, this works, but on ARM the result is evaluated (and thus "ABC" is printed) _before_ the "Output is: " string is printed, so the output gets messed up.

The attached patch fixes this.

(I still get three failures, but I'll open another bug for them as it's a separate issue.)

Comment 1

[Jacob Bramley [:jbramley]]

The other failures are already covered by bug 525412.

Comment 2

[Graydon Hoare :graydon]

Comment on attachment 409695 [details] [diff] [review]
Fix lirasm.cpp so that it evaluates rint() and rfloat() before it starts to print the result.

Hilarious. iostreams kicks us in the shins again. Thanks.

Comment 3

[Graydon Hoare :graydon]

http://hg.mozilla.org/projects/nanojit-central/rev/9088798ea4a9

Comment 4

[Luke Wagner [:luke]]

(In reply to comment #2)
> (From update of attachment 409695 [details] [diff] [review])
> Hilarious. iostreams kicks us in the shins again. Thanks.

It would seem that the non-determinism allowed by C sequence points is to blame, not our beleaguered type safe friend iostreams.  The same flexibility to evaluate argument expressions would have been present if printf was used.

Comment 5

[Robert Sayre]

http://hg.mozilla.org/mozilla-central/rev/651db313edf2

Comment 6

[Samuel Sidler (old account; do not CC)]

These bugs landed after b4 was cut. Moving flag out.