This looks nice! We should probably also have a test for the case where a user specifies an END before a BEGIN tag.

Thanks for the review, I addressed your comments:

• Removed the reference, and pass the AST node as a pointer.
• Made a few decls auto

LGTM, I'll let @jhenderson weigh-in to see what he says.

LGTM, as long as @rupprecht's comment is addressed.

In D53832#1450240, @troyj wrote:

Could you simply rebase the patch, or acknowledge the new LLVM license, as your patch was written prior to the license change.

My organization (Cray Inc) already signed the new license agreement, so I'm not sure what you need here. I can write that I acknowledge the new LLVM license, if that helps.

In D53832#1443179, @troyj wrote:

I do not have commit access. I probably should request it at some point, but I'm fine with someone else merging this for me.

LGTM!

In D53832#1439874, @troyj wrote:

This revision is now accepted and ready to land.Dec 3 2018, 12:55 PM

Been about 3.5 months since accepted. Can this be committed?

Makes sense. LGTM.

Thanks for the reviews everyone. This update addresses the feedback, all of which was incorporated.

LGTM

I like this and think users will find it very helpful. The changes seem sensible to me; however, I had a few nits... mostly stylistic things, no big deals. Anyways, I'll mark this patch as accept, as long as you cover the suggestions made by @lebedev.ri.

• Fixed the test case (removed underscore typo, and extraneous shstrtab test).
• Emit ANDROID_REL ANDROID_RELA instead of section offsets.
• Modified the range check, allowing us to remove some now redundant code for printing the lower-bound section names.

In D58701#1411546, @Bigcheese wrote:

lgtm

Should the +0 be +0x0 so that it's always a hex digit? Or is the +0 what the binutils already output?

• Added a more complicated test.
• Made a few variables 'auto' instead of their iterator type.

LGTM. Thanks for adding these.

In D58416#1403227, @compnerd wrote:

Previously we would have emitted Foo [[:space:]] . right?

In D57149#1388361, @uabelho wrote:

In D57149#1387325, @mattd wrote:

Hi @uabelho

The test case in your patch definitely causes a problem, but that's because the -start-before/-stop-after is used. If I run that same test through llc, the UnreachableBlockElim pass kicks-in and drops the dead code.
I took a look at the code in your Janurary 24th comment. I was unable to get the same dead block to surface, but most likely I had the wrong flag set. How did you build the sample in that comment?

It can be reproduced with the full llc -O1 pipe on current trunk using:

llc -O1 -o - pr40436.ll

pr40436.ll825 BDownload

LGTM. I had a few suggestions but nothing looks wrong to me.

I've updated the patch to replicate GNU's c++filt behavior. This patch now only splits strings that are passed via stdin. It does not split strings that are passed as arguments to llvm-cxxfilt via command line.

The test case in your patch definitely causes a problem, but that's because the -start-before/-stop-after is used. If I run that same test through llc, the UnreachableBlockElim pass kicks-in and drops the dead code.
I took a look at the code in your Janurary 24th comment. I was unable to get the same dead block to surface, but most likely I had the wrong flag set. How did you build the sample in that comment?

In D57700#1384786, @jhenderson wrote:

Hello Davide,
The present format is good enough to convey the information. It is however not sufficient or sometimes difficult for present users of GNU tools (which are quite ubiquitous) to make a
switch as it is complete redesign of the post processing tools they may have based on.

-Hemant

It's not immediately clear from the original introduction of GNU style whether complete output identity was the original goal, or just similarity. Similarly, it's not clear from my previously attempt to gather information as to users expectations. I've subscribed @khemant in case he's got any further comments, since he made the original change. My reading of rL260430 which commits it is that it is intended to be like GNU readelf, but doesn't necessarily have to be identical. I'm open to further discussion on this (and I actually have a BoF on this subject at the next Euro LLVM, but that's still a couple of months away).

• Update two tests to use FileChecks --implicit-check-not
• Remove some cruft that, I think, was in an earlier version of this patch.
• Update the CHECK-COUNT test to simply perform a CHECK followed by a CHECK-NOT.

In D57105#1378441, @jhenderson wrote:

Okay, the change as-is LGTM, assuming that the following has been done:

I think before putting this in as is, it's probably worth reviewing all the call sites and seeing if there is anywhere where returning the section name would be harmful. I suspect it's probably mostly fine, given the lack of test failures.

A few things to look out for:

1. Some code paths may already do the extra work needed to look up the section name, if the symbol doesn't have a name. These should be tidied up to not do this any more (it can be a later change).
2. Some code paths may always print the name returned, but this may be undesirable for certain situations (I can't think of any off the top of my head however). This wouldn't be picked up by tests necessarily, due to the way FileCheck tests tend to be written.
3. Some code may be using the name in some kind of map, if non-empty. This could cause issues e.g. with relocations in LLD, when attempting to map references to symbols via names. In practice, it shouldn't be an issue, but we should double check.

• I took a hint and decided to more the decision to print the section mapping with the program headers down into the dumper classes. This is similar to how the printSymbols and printDynamicSymbols are handled.

In D57365#1376848, @jhenderson wrote:

It occurred to me that if we were to implement section mapping for other output styles, e.g. LLVM, or formats (I don't know if that concept applies to COFF/MachO/Wasm etc mind you!), then we probably don't want to print the section mapping when program-headers is specified, except in GNU output style. This implies that the decision on whether to print the section mapping should be down to the dumper layer, not the command-line handling layer. This is very similar to the situation with --symbols/--dyn-symbols for GNU output style, and I suspect the solution should be essentially the same as rL351960.

• Removed a test
• Replaced the uses of grep in the tests with CHECK-NOT and CHECK-COUNT

• Update a comment, it was bothering me.