1614375 - ubsan pointer-overflow errors with clang 10

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[(Away)]

There are various instances of this error showing up in try pushes, but the most prevalent one (which makes pretty much every linux asan test orange) is:

xmlparse.c:1942:24: runtime error: applying zero offset to null pointer

https://searchfox.org/mozilla-central/rev/a1592902acabf9bded973067133baaac1457f3d3/parser/expat/lib/xmlparse.c#1942

Comment 1

[(Away)]

The amount of pointer-overflow diagnostics has increased considerably in LLVM 10: https://github.com/llvm/llvm-project/commit/536b0ee40ab97f2878dc124a321cf9108ee3d233

Comment 2

[Jeff Walden [:Waldo]]

Slightly tangentially, the error mentioned in comment 0 is specified to be UB in some versions of C++, but C++17 specifically relaxed things in this regard: "If the value 0 is added to or subtracted from a null pointer value, the result is a null pointer value." C++17 [expr.add]p7.

But xmlparse.c is obviously not C++, and neither C11 nor C18 seem to include the language C++17 has to make clear that adding zero to a null pointer results in a null pointer.

Comment 3

[(Away)]

Yeah, the commit message from clang says:

To make things more fun, in C (6.5.6p8), applying *any* offset to null pointer
is undefined, although Clang front-end pessimizes the code by not lowering
that info, so this UB is "harmless".

Comment 4

[(Away)]

Attachment: UBSan pointer-overflow suppressions for clang-10

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/aa44cf2c1455
UBSan pointer-overflow suppressions for clang-10 r=tsmith

Comment 6

[Cosmin Sabou [:CosminS]]

https://hg.mozilla.org/mozilla-central/rev/aa44cf2c1455

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/8311d3bf80c0
bustage fix, add a couple more suppressions for straggler failures

Comment 8

[Andreea Pavel [:apavel]]

https://hg.mozilla.org/mozilla-central/rev/8311d3bf80c0