This fixes pr34301.

As the bug points out, we want to keep some relocations with undefined
weak symbols. This means that we cannot always claim that these
symbols are not preemptible as we do now.

Unfortunately, we cannot also just always claim that they are
preemptible. Doing so would, for example, cause us to try to create a
plt entry when we don't even have a dynamic symbol table.

What almost works is to say that weak undefined symbols are
preemptible if and only if we have a dynamic symbol table. Almost
because we don't want to fail the build trying to create a copy
relocation to a weak undefined.

llvm-svn: 313372

Patch removes one of OutputSectionFactory::addInputSec methods.
That allows to simplify reporting of discarded sections and
should help to D37561.

Differential revision: https://reviews.llvm.org/D37735

llvm-svn: 313361

This patch is still breaking several multi-stage compiler-rt bots.
I already know what the fix is, but I want to get the bots green
for now and then try re-applying in the morning.

llvm-svn: 313335

_gp points to a position in the file, so it is not really absolute. It
is also simpler to not force it to be absolute, so if there is no
strong ABI requirement we should not do it.

llvm-svn: 313333

llvm-svn: 313330

llvm-svn: 313289

GNU ld manual says that multi-letter long option can be prefixed with
either -- or -. Therefore, we should accept not only --subsystem but
also -subsystem, for example.

There is one exception. If an option starts with "o", it should only be
prefixed with -- to avoid ambiguity with -o<filename> option.

Differential Revision: https://reviews.llvm.org/D37825

llvm-svn: 313286

Differential Revision: https://reviews.llvm.org/D37823

llvm-svn: 313285

This patch simplifies LLVM's lit infrastructure by enforcing an ordering
that a site config is always run before a source-tree config.

A significant amount of the complexity from lit config files arises from
the fact that inside of a source-tree config file, we don't yet know if
the site config has been run.  However it is *always* required to run
a site config first, because it passes various variables down through
CMake that the main config depends on.  As a result, every config
file has to do a bunch of magic to try to reverse-engineer the location
of the site config file if they detect (heuristically) that the site
config file has not yet been run.

This patch solves the problem by emitting a mapping from source tree
config file to binary tree site config file in llvm-lit.py. Then, during
discovery when we find a config file, we check to see if we have a
target mapping for it, and if so we use that instead.

This mechanism is generic enough that it does not affect external users
of lit. They will just not have a config mapping defined, and everything
will work as normal.

On the other hand, for us it allows us to make many simplifications:

* We are guaranteed that a site config will be executed first
* Inside of a main config, we no longer have to assume that attributes
  might not be present and use getattr everywhere.
* We no longer have to pass parameters such as --param llvm_site_config=<path>
  on the command line.
* It is future-proof, meaning you don't have to edit llvm-lit.in to add
  support for new projects.
* All of the duplicated logic of trying various fallback mechanisms of
  finding a site config from the main config are now gone.

One potentially noteworthy thing that was required to implement this
change is that whereas the ninja check targets previously used the first
method to spawn lit, they now use the second. In particular, you can no
longer run lit.py against the source tree while specifying the various
`foo_site_config=<path>` parameters.  Instead, you need to run
llvm-lit.py.

Differential Revision: https://reviews.llvm.org/D37756

llvm-svn: 313270

llvm-svn: 313204

The isShared case was handled by the isInCurrentDSO check.

llvm-svn: 313190

llvm-svn: 313188

Arg instances can be claimed. After claimed, its `isClaimed` function
returns true. We do not use that notion in lld, so using NoClaim
versions of functions is just confusing. This patch is to just use
hasArg instead of hasArgNoClaim.

llvm-svn: 313187

llvm-svn: 313184

There are no alises handled by this switch, but getUnaliasesdOption is
preferred way of doing this. This is also consistent with ELF and COFF.

llvm-svn: 313180

llvm-svn: 313178

This is how the flag is documented in GNU binutils ld; -Bstatic
only applies to -l options after it, until the next -Bdynamic.

Differential Revision: https://reviews.llvm.org/D37794

llvm-svn: 313175

In MinGW configurations (GCC, or clang with a *-windows-gnu target),
the -export directives in the object file contains the undecorated
symbol name, while it is decorated in MSVC configurations. (On the
command line, link.exe takes an undecorated symbol name for the
-export argument though.)

Differential Revision: https://reviews.llvm.org/D37772

llvm-svn: 313174

This reduces the size of SharedSymbol which in turn reduces the size
of Symbol from 88 to 80 bytes.

llvm-svn: 313154

llvm-svn: 313129

There is no need to scan over all input sections for relocatable output.
As we do not process or scan relocations anyways.
Patch moves check for Config->Relocatable out to avoid that and also removes
excessive check for isa<EhInputSection> from first for loop. 
It is excessive because we handle all of them in a second for loop below.
That all allowed to simplify code.

Differential revision: https://reviews.llvm.org/D37746

llvm-svn: 313127

Differential Revision: https://reviews.llvm.org/D37761

llvm-svn: 313125

Differential Revision: https://reviews.llvm.org/D37769

llvm-svn: 313124

This fixes PR31824.

Differential Revision: https://reviews.llvm.org/D37709

llvm-svn: 313123

This should fix the lto bootstrap.

It is somewhat hard to remember about lazy symbols deep down in the
link. It might be worth it replacing them with undefined symbols once
we are done adding files.

llvm-svn: 313103

llvm-svn: 313095

llvm-svn: 313094

https://reviews.llvm.org/rL312796 meant that references to garbage collected common symbols would cause a segfault.

This change fixes the behaviour for references to stripped common symbols.

Differential Revision: https://reviews.llvm.org/D37718

llvm-svn: 313086

It is not needed since it is always 0.

llvm-svn: 313076

On i386, the --entry parameter to GNU ld is supposed to be a decorated
symbol name, while it is an undecorated name in link.exe.

Differential Revision: https://reviews.llvm.org/D37710

llvm-svn: 313066

llvm-svn: 313051

Given our representation of symbols as section + offset, we can
finalize symbols like _end as soon as the section is known.

llvm-svn: 313048

Differential Revision: https://reviews.llvm.org/D37708

llvm-svn: 313045

For now LLD does not setup the least-significant bit for microMIPS
symbols. llvm-objdump does not like that. In attempt to fix
sanitizer-x86_64-linux-fast build-bot let's temporarily check the raw
binary file content.

llvm-svn: 313040

The patch implements initial support of microMIPS code linking:
  - Handle microMIPS specific relocations.
  - Emit both R1-R5 and R6 microMIPS PLT records.

For now linking mixed set of regular and microMIPS object files is not
supported. Also the patch does not handle (setup and clear) the
least-significant bit of an address which is utilized as the ISA mode
bit and allows to make jump between regular and microMIPS code without
any thunks.

Differential revision: https://reviews.llvm.org/D37335

llvm-svn: 313028

Replace OutputSection *Cmd to OutputSection *OS. The Commands vector was
moved to OutputSection but the names of the variables were not. This patch
changes the names to match.

Differential Revision: https://reviews.llvm.org/D37627

llvm-svn: 313015

This fixes two more cases where we were aligning the offset in a
section, instead of the final address.

llvm-svn: 312983

llvm-svn: 312981

When given

foobar = ALIGN(., 0x100);

my expectation from what the manual says is that the final address of
foobar will be aligned. It seems that bfd aligns the offset in the
section, which causes some odd results if the section is not 0x100
aligned. Gold aligns the address.

This changes lld to align the final address.

llvm-svn: 312979

llvm-svn: 312976