Update size computation + test case involving null byte inside the format string.

Take into account some of the reviews

Add test case

In D72869#1824951, @efriedma wrote:

This is very hacky,

The simplest one I can think of is to make __warn_memset_zero_len a recognized builtin that generates no code.

+1, simple and easy.

@akhuang can you confirm this fixes your issue? I've tested locally and it works fine on my side.

In D71082#1824563, @rnk wrote:

The simplest one I can think of is to make __warn_memset_zero_len a recognized builtin that generates no code.

@erik.pilkington up?

@MaskRay no example yet, I'd like your opinion on the approach first. It has the advantage of being non intrusive to the code base...

@craig.topper up?

@akhuang : I've reported https://sourceware.org/bugzilla/show_bug.cgi?id=25399 to glibc. According to me this is more of a glibc problem than this patch, and as we have a workaround with -fgnuc-version=5.0 I think we could apply the patch back.

In D71082#1822925, @akhuang wrote:

This caused a linker error in chromium:

ld.lld: error: undefined symbol: __warn_memset_zero_len

Apparently now that the glibc memset is being used, __warn_memset_zero_len gets called from libc++ code (https://github.com/llvm/llvm-project/blob/b72a8c65e4e34779b6bc9e466203f553f5294486/libcxx/include/__bit_reference#L371).

Obsoleted by https://reviews.llvm.org/D72724, closing. Thanks @Romain-Geissler-1A for triggering that patch!

@MaskRay: is that okay with you if I start implementing -fsemantic-interposition based on dso_preemptable?

LGTM

(back on that one) The default in clang is implicitly -fno-semantic-interposition. I think we can safely support it, and either warn or error on -fsemantic-interposition. @hfinkel does that seem correct to you?

@lebedev.ri description updated.

• update warn_fe_stack_clash_protection_inline_asm location.

Take review into account.

@erik.pilkington up?

With this patch applied, I've validated on OSX and Linux that load_extension.ll passes the validation with Polly linked both statically and dynamically.

• Fix OSX build.

@sylvestre.ledru : I'd like a confirmation from @efriedma first!

@xbolva00 : gentle reminder :-)

I like the idea but

In D72327#1808263, @fhahn wrote:

Does this fix llvm/test/Feature/load_extension.ll on macOS by any chance?

• Test and handle the combination of LLVM_BUILD_EXAMPLE=OFF + LLVM_BYE_LINK_INTO_TOOLS=ON

I've tested that with LLVM_BYE_LINK_INTO_TOOLS ON/OFF, LLVM_BUILD_EXAMPLES ON/OFF LLVM_POLLY_LINK_INTO_TOOLS ON/OFF and it seems okay.

In D61446#1805413, @svenvh wrote:

This change seems to be causing a problem with the nightly packages from apt.llvm.org.

Correct, there's a missing dependency, working on it.

• correctly declare example dependency

In rG0d275431a3abc96fdee3e09afdc84e59df0e1d3b#871635, @RKSimon wrote:

@serge-sans-paille http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-ubuntu-fast/ is still failing - please can you take a look?

Patch rebased, validation ongoing. I'll merge that once I'm back from holidays, don't feel like commiting that 3h before holidays.

Support %% and %c

• Prevent assert upon wide string format
• More test cases

Take remarks into account:

• support sprintf when the buffer has a known static size
• use llvm::Optional
• some extra minor tweaks.

@efriedma validation passes without the checks you were pointed out, see https://github.com/serge-sans-paille/llvm-project/pull/4/checks
Thanks for making the code simpler!

Take comment into account.

@xbolva00 : validation is ok: https://github.com/serge-sans-paille/llvm-project/pull/5/checks

Added test cases

[[edited the example to reflect actual state]]

In D71374#1783245, @xbolva00 wrote:

LLVM already converts mempcpy to memcpy..

In D71374#1783032, @Jim wrote:

I am curious what is difference of code generation after applying your changes?

Handle comments.

Improve test case portability.

@Jim obviously :-) Thanks for spotting that.

@george.burgess.iv : take into account reviews, extra testing and function renaming.

In D71374#1780402, @serge-sans-paille wrote:

On going validation: https://github.com/serge-sans-paille/llvm-project/pull/5/checks

On going validation: https://github.com/serge-sans-paille/llvm-project/pull/5/checks