pawosm01 added a comment.
You're right, this patch smuggles elimination of two dead functions. I've
prepared alternative version of this patch which does not do that.
Repository:
rL LLVM
http://reviews.llvm.org/D16929
___
cfe-commits mailing lis
pawosm-arm wrote:
The problem with `-fallow-argument-mismatch` has already been reported to the
MPICH project maintainers, https://github.com/pmodels/mpich/issues/6869
therefore I'm going to close this PR.
https://github.com/llvm/llvm-project/pull/91611
___
https://github.com/pawosm-arm closed
https://github.com/llvm/llvm-project/pull/91611
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/pawosm-arm approved this pull request.
This is great, thanks for contributing it.
https://github.com/llvm/llvm-project/pull/100343
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listin
https://github.com/pawosm-arm approved this pull request.
The -fno-fortran-main flag should disappear before it finds any widespread use.
https://github.com/llvm/llvm-project/pull/89938
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://l
https://github.com/pawosm-arm commented:
As I'm not seeing how linker is invoked, I just need to ask (for the sake of
completeness) will it also prevent addition of `-lomp` to the linker's command
line?
https://github.com/llvm/llvm-project/pull/107087
__
https://github.com/pawosm-arm approved this pull request.
https://github.com/llvm/llvm-project/pull/107087
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/pawosm-arm created
https://github.com/llvm/llvm-project/pull/91611
Many autotools-utilizing projects (mpich among them) fail to complete the
`configure` step since it tries to invoke the (unknown to them) Fortran
compiler always with the `-fallow-argument-mismatch` flag.
As
pawosm-arm wrote:
>
> It sounds like an issue with the `mpich` build script rather than LLVM Flang.
> Why would `mpich` and other projects use this flag to begin with? Do other
> Fortran compilers support it?
>
I don't know why they do that, I suppose they think they know better. I don't
sus
pawosm-arm wrote:
The number of C++ workloads failing to build after #90152 is rather impressive,
even this PR do not help. Is there a plan to address that?
https://github.com/llvm/llvm-project/pull/91503
___
cfe-commits mailing list
cfe-commits@lists
pawosm-arm wrote:
I'm sorry, I didn't notice that there is more discussion after #90152 covering
also my issue at hand. The obvious advice I can get from those is the 'this is
not valid c++, fix your code' suggestion. Trouble starts when there is a lot of
it to fix, even if it mostly involve d
https://github.com/pawosm-arm created
https://github.com/llvm/llvm-project/pull/114837
When linking with other runtimes (OpenMP, sanitizers), the
addArchSpecificRPath() is being called. The same thing should happen when
linking with the Fortran runtime, this will improve user experience massiv
pawosm-arm wrote:
> We should probably also handle `no-infs-fp-math` to save the next person the
> trouble.
I'd leave it to a separate PR, as this one has a specific story behind the
proposed change, anything extra would go beyond it.
https://github.com/llvm/llvm-project/pull/114271
_
pawosm-arm wrote:
> > Yet we see a value in the change proposed here too hence a good reason for
> > merging it.
>
> I've clearly misunderstood the context of the rebase. What is the value in
> merging this change if the original issue has already been resolved? I guess
> this fix is simpler
pawosm-arm wrote:
> `addSanitizerRuntime` and `addOpenMPRuntime` are already calling
> `addArchSpecificRPath`. It it a problem if rpath is added multiple times?
>
I've been testing it with and without -fopenmp and didn't observe any problem
> Compiler-rt libs (`libclang_rt.*.a`) are added as
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/114271
>From a426e17ed3c3cad6b3c813bd996e649a2e544591 Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Wed, 30 Oct 2024 14:30:08 +
Subject: [PATCH] [Transforms][Utils][PromoteMem2Reg] Propagate nnan and ni
pawosm-arm wrote:
> > I've rebased my PR so this branch could also contain phase-ordering test
> > for vectorizing predicated selects introduced by commit
> > [577c7dd](https://github.com/llvm/llvm-project/commit/577c7dd7cc4c5a9f62f9654cfa30ee9d55709426)
>
> Do I understand correctly that this
pawosm-arm wrote:
If there is no objections by others, I'll merge it some time tomorrow.
https://github.com/llvm/llvm-project/pull/114837
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commi
pawosm-arm wrote:
> `addSanitizerRuntime` and `addOpenMPRuntime` are already calling
> `addArchSpecificRPath`. It it a problem if rpath is added multiple times?
For a reference, in case of Clang and C code utilizing OpenMP, -rpath is given
at least twice: `"/usr/bin/ld" "-rpath"
"/opt/ll
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/114271
>From 45a02d4dd6102bf825943e738a15adf64361f976 Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Wed, 30 Oct 2024 14:30:08 +
Subject: [PATCH] [Transforms][Utils][PromoteMem2Reg] Propagate nnan and ni
pawosm-arm wrote:
There was a change of plans, so I don't need it now. But I expect the subject
will return sooner or later raised by someone else who stumble upon this just
like me.
Closing it for now.
https://github.com/llvm/llvm-project/pull/110598
https://github.com/pawosm-arm closed
https://github.com/llvm/llvm-project/pull/110598
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
pawosm-arm wrote:
Ah, so it isn't failing only for me. Apparently, it blows away all the
environments where ORC is deliberately disabled.
https://github.com/llvm/llvm-project/pull/110418
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://
pawosm-arm wrote:
> I don't think it is correct to say that the option was ignored. It may still
> be read by the compiler driver (which generates the link line). In fact it
> probably _should_ be processed by the compiler driver in order to set the C++
> standard library correctly - otherwise
@@ -0,0 +1,48 @@
+// REQUIRES: system-linux
pawosm-arm wrote:
As I stated in my other comment, the `-###` output is not affected by the
`-stdlib=` flag, so a test which looks at its output would not provide a great
value. The integration test from the other han
pawosm-arm wrote:
> If this is completely ignored and does not effect the link line, are you sure
> the resulting binary actually uses the right c++ library?
>
> Edit: flang will not link a C++ library. It is not a C++ compiler.
As we discussed it offline, in case of this project it does not m
pawosm-arm wrote:
> Note that the test will fail when `LLVM_DEFAULT_TARGET_TRIPLE` is set to
> anything else than the host platform. Because of this. Clang tests never
> invoke the linker, but check the output of `-###`. However, Flang already has
> tests with the same issue.
I wonder how the
pawosm-arm wrote:
> (and there isn't even a warning that the option is unused)
>
> @pawosm-arm Wasn't this ever an issue? Should those to be consistent?
Interesting, clang would do the same as flang-new patched with my patch. Maybe
addition of the warning would be a good idea, if it isn't an o
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/115163
>From 2c4b767437c55245391eb15da94a97b34a63f834 Mon Sep 17 00:00:00 2001
From: Paul Osmialowski
Date: Tue, 5 Nov 2024 16:38:22 +
Subject: [PATCH] [clang][Driver] Add the ability to specify that RPATH shoul
pawosm-arm wrote:
We worked out this patch just because we dropped the idea of adding
configuration file to our toolchain and for no other reason...
https://github.com/llvm/llvm-project/pull/115163
___
cfe-commits mailing list
cfe-commits@lists.llvm.o
pawosm-arm wrote:
> Some distributions (e.g. Fedora) have policies against DT_RUNPATH for system
> packages.
For as long as I remember, distros (and Fedora is a good example of this
phenomenon) were patching the software packages heavily, so adding one more to
ensure their well established po
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/114271
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/pawosm-arm created
https://github.com/llvm/llvm-project/pull/115163
The `-frtlib-add-rpath` is very convenient when linking against various
runtimes (OpenMP, Fortran, sanitizers), so much so we could argue that this
should be a default behavior.
This patch adds the ability
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/115286
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
pawosm-arm wrote:
> For what it’s worth, as a user I’m always surprised that this _isn’t_ the
> default. It’s quite annoying to eg have to set LD_LIBRARY_PATH when using
> openmp with clang when they’re built alongside each other. Is there a
> specific reason this isn’t the default or is it ju
https://github.com/pawosm-arm created
https://github.com/llvm/llvm-project/pull/115286
The `-frtlib-add-rpath` is very convenient when linking against various
runtimes (OpenMP, Fortran, sanitizers), so much so we could argue that this
should be a default behavior.
This patch makes adding RPAT
pawosm-arm wrote:
@paulwalker-arm Can you share your rationale here?
https://github.com/llvm/llvm-project/pull/115286
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -0,0 +1,34 @@
+! REQUIRES: x86-registered-target
pawosm-arm wrote:
removed those requirements
https://github.com/llvm/llvm-project/pull/115163
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.ll
pawosm-arm wrote:
As I prepared a piece of code which makes the `-l` flags being treated properly
when read from a config file (so it won't result in unresolved symbols e.g.
when used with `-static`), I realized that going the config file way still
creates a problem: with `-fveclib=ArmPL -lama
pawosm-arm wrote:
No, config files today ARE NOT sufficient to do that. There is no way to attach
a role to the flags in clang's config file, all flags are being treated the
same. So putting -lamath in there will put it before object files that would
make use of it, it will result in unresolve
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/116432
>From 5614595f48dc2a6b4087c992c108874ca6c787ee Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Fri, 15 Nov 2024 15:22:21 +
Subject: [PATCH] [clang][driver] When -fveclib=ArmPL flag is in use, alway
pawosm-arm wrote:
> We probably have to provide a link to some documentation for libamath. May be
> something like the following one.
> https://developer.arm.com/documentation/102574/2410/Optimized-math-routines---libamath
I'm always anxious to provide a link that contains numeric ID (like the
pawosm-arm wrote:
I've extended it with Windows and Darwin.
https://github.com/llvm/llvm-project/pull/116432
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/116432
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/pawosm-arm closed
https://github.com/llvm/llvm-project/pull/115286
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/116432
>From 392192f186be03186cd53e2f94b56a21c5db27aa Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Fri, 15 Nov 2024 15:22:21 +
Subject: [PATCH] [clang][driver] When -fveclib=ArmPL flag is in use, alway
https://github.com/pawosm-arm closed
https://github.com/llvm/llvm-project/pull/115163
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -490,6 +490,16 @@ void tools::AddLinkerInputs(const ToolChain &TC, const
InputInfoList &Inputs,
else
A.renderAsInput(Args, CmdArgs);
}
+ if (const Arg *A = Args.getLastArg(options::OPT_fveclib)) {
+if (A->getNumValues() == 1) {
+ StringRef V = A->getVa
@@ -490,6 +490,16 @@ void tools::AddLinkerInputs(const ToolChain &TC, const
InputInfoList &Inputs,
else
A.renderAsInput(Args, CmdArgs);
}
+ if (const Arg *A = Args.getLastArg(options::OPT_fveclib)) {
+if (A->getNumValues() == 1) {
+ StringRef V = A->getVa
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/116432
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -61,3 +61,24 @@
! CHECK-TWO-CONFIGS-NEXT: Configuration file:
{{.*}}Inputs{{.}}config2{{.}}config-4.cfg
! CHECK-TWO-CONFIGS: -ffp-contract=fast
! CHECK-TWO-CONFIGS: -O3
+
+!--- The -l flags should be moved to the end of input list and appear only
when linking.
+! RUN: %fla
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/117573
>From d5a424405ef75b566167ce1686912a3d83949c48 Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Mon, 25 Nov 2024 14:46:55 +
Subject: [PATCH] [clang][driver] Special care for -l and -Wl, flags in con
@@ -1250,6 +1273,7 @@ Compilation *Driver::BuildCompilation(ArrayRef ArgList) {
if (!ContainsError)
ContainsError = loadConfigFiles();
bool HasConfigFile = !ContainsError && (CfgOptions.get() != nullptr);
+ bool HasConfigLinkerIn = !ContainsError && (CfgLinkerInputs.ge
@@ -82,3 +82,24 @@
// CHECK-TWO-CONFIGS: -isysroot
// CHECK-TWO-CONFIGS-SAME: /opt/data
// CHECK-TWO-CONFIGS-SAME: -Wall
+
+//--- The -l flags should be moved to the end of input list and appear only
when linking.
+// RUN: %clang --target=aarch64-unknown-linux-gnu --config
%S
@@ -61,3 +61,24 @@
! CHECK-TWO-CONFIGS-NEXT: Configuration file:
{{.*}}Inputs{{.}}config2{{.}}config-4.cfg
! CHECK-TWO-CONFIGS: -ffp-contract=fast
! CHECK-TWO-CONFIGS: -O3
+
+!--- The -l flags should be moved to the end of input list and appear only
when linking.
+! RUN: %fla
pawosm-arm wrote:
> I took a look and I think it may be too awkward to do, as we'd want to run
> e.g. readelf afterwards. But an example is
> `lld/test/ELF/as-needed-not-in-regular.s`. The idea being: if
> `-Wl,--as-needed` is in the config file, do we correctly prune an unnecessary
> library
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/117573
>From bc66202680198f0e5d1289c62f79ca5799e6f2ac Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Mon, 25 Nov 2024 14:46:55 +
Subject: [PATCH] [clang][driver] Special care for -l and -Wl, flags in con
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -0,0 +1 @@
+-ffast-math -Wl,--as-needed | -lm -Wl,-Bstatic -lhappy -Wl,-Bdynamic
pawosm-arm wrote:
Since for ` llvm::cl::ExpansionContext::readConfigFile()` a newline is just a
separator, this can't be achieved by doing an enhancement in a code very
distant
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -0,0 +1 @@
+-ffast-math -Wl,--as-needed | -lm -Wl,-Bstatic -lhappy -Wl,-Bdynamic
pawosm-arm wrote:
We could prefix those linker flags with some character, e.g. '@', it would work
like that:
```
-ffast-math @-lm
-Wl,--as-needed @-Wl,-Bstatic -lhappy
@-Wl,-Bdy
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -0,0 +1 @@
+-ffast-math -Wl,--as-needed | -lm -Wl,-Bstatic -lhappy -Wl,-Bdynamic
pawosm-arm wrote:
mid-air collision. yeah, either '$' or '@' would do :)
https://github.com/llvm/llvm-project/pull/117573
___
cfe-com
@@ -0,0 +1 @@
+-ffast-math -Wl,--as-needed | -lm -Wl,-Bstatic -lhappy -Wl,-Bdynamic
pawosm-arm wrote:
Seems neither '$' nor '@' is the best choice, '$' looks like shell variable
expansion, so I'll do `^`, e.g., '^-Wl,-Bstatic`.
https://github.com/llvm/llvm-pro
@@ -0,0 +1 @@
+-ffast-math -Wl,--as-needed | -lm -Wl,-Bstatic -lhappy -Wl,-Bdynamic
pawosm-arm wrote:
Done.
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/116432
>From 0906f315571cfa093889077c5ec155cae3d174b6 Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Fri, 15 Nov 2024 15:22:21 +
Subject: [PATCH] [clang][driver] When -fveclib=ArmPL flag is in use, alway
pawosm-arm wrote:
> Typo in PR title `flang` -> `flag`
Corrected both the title and the commit message. Thanks for spotting this!
https://github.com/llvm/llvm-project/pull/116432
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.ll
https://github.com/pawosm-arm created
https://github.com/llvm/llvm-project/pull/116432
Using `-fveclib=ArmPL` without `-lamath` likely effects in the link-time errors.
>From 991304ba11c9852964a0f0bdcde5277d47e29c7c Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Fri, 15 Nov 2024 15:22:2
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/116432
>From 865831ba2fb4778e8a1c25402b1cff70abd44d6b Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Fri, 15 Nov 2024 15:22:21 +
Subject: [PATCH] [clang][driver] When -fveclib=ArmPL flag is in use, alway
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/116432
>From dde726aa72cd23a5a824846e6b65583d1274d18c Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Fri, 15 Nov 2024 15:22:21 +
Subject: [PATCH] [clang][driver] When -fveclib=ArmPL flag is in use, alway
https://github.com/pawosm-arm created
https://github.com/llvm/llvm-project/pull/117573
Currently, if a -l flag is added into a config file (e.g. clang.cfg), it is
situated before any object file in the effective command line. If the library
requested by given -l flag is static, its symbols wil
@@ -61,3 +61,22 @@
! CHECK-TWO-CONFIGS-NEXT: Configuration file:
{{.*}}Inputs{{.}}config2{{.}}config-4.cfg
! CHECK-TWO-CONFIGS: -ffp-contract=fast
! CHECK-TWO-CONFIGS: -O3
+
+!--- The -l flags should be moved to the end of input list and appear only
when linking.
+! RUN: %fla
pawosm-arm wrote:
> I think it makes sense to handle linker options differently so I'm in favour
> of this change in principle.
>
> Am I right in thinking that if the config file puts things last, the `-l`
> options provided by users will come before the config file ones, and unlike
> other o
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -0,0 +1 @@
+-ffast-math -lm -lhappy
pawosm-arm wrote:
yes
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/117573
>From 2acf432b48c275492d9c5d012e85da36750c9299 Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Mon, 25 Nov 2024 14:46:55 +
Subject: [PATCH] [clang][driver] Special care for -l and -Wl, flags in con
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/117573
>From e5769df6180f6b89ad2c494e74d3e4dc9d88dec8 Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Mon, 25 Nov 2024 14:46:55 +
Subject: [PATCH] [clang][driver] Special care for -l and -Wl, flags in con
@@ -82,3 +82,29 @@
// CHECK-TWO-CONFIGS: -isysroot
// CHECK-TWO-CONFIGS-SAME: /opt/data
// CHECK-TWO-CONFIGS-SAME: -Wall
+
+//--- The linker input flags should be moved to the end of input list and
appear only when linking.
+// RUN: %clang --target=aarch64-unknown-linux-gnu --
@@ -61,3 +61,29 @@
! CHECK-TWO-CONFIGS-NEXT: Configuration file:
{{.*}}Inputs{{.}}config2{{.}}config-4.cfg
! CHECK-TWO-CONFIGS: -ffp-contract=fast
! CHECK-TWO-CONFIGS: -O3
+
+!--- The linker input flags should be moved to the end of input list and
appear only when linking.
+!
pawosm-arm wrote:
> This seems OK to me, although I'd feel more comfortable if we had a test
> which relied on `--Wl,--as-needed` too.
Could you provide or describe some example?
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits maili
@@ -1062,6 +1062,16 @@ bool Driver::readConfigFile(StringRef FileName,
for (Arg *A : *NewOptions)
A->claim();
+ std::unique_ptr NewLinkerIns =
std::make_unique();
pawosm-arm wrote:
changed and added a comment
https://github.com/llvm/llvm-project/pul
@@ -61,3 +61,29 @@
! CHECK-TWO-CONFIGS-NEXT: Configuration file:
{{.*}}Inputs{{.}}config2{{.}}config-4.cfg
! CHECK-TWO-CONFIGS: -ffp-contract=fast
! CHECK-TWO-CONFIGS: -O3
+
+!--- The linker input flags should be moved to the end of input list and
appear only when linking.
+!
https://github.com/pawosm-arm edited
https://github.com/llvm/llvm-project/pull/117573
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -61,3 +61,29 @@
! CHECK-TWO-CONFIGS-NEXT: Configuration file:
{{.*}}Inputs{{.}}config2{{.}}config-4.cfg
! CHECK-TWO-CONFIGS: -ffp-contract=fast
! CHECK-TWO-CONFIGS: -O3
+
+!--- The linker input flags should be moved to the end of input list and
appear only when linking.
+!
@@ -1073,6 +1086,19 @@ bool Driver::readConfigFile(StringRef FileName,
appendOneArg(*CfgOptions, Opt, BaseArg);
}
}
+
+ if (!CfgLinkerInputs)
+CfgLinkerInputs = std::move(NewLinkerIns);
+ else {
+// If this is a subsequent config file, append options to th
@@ -1062,6 +1062,19 @@ bool Driver::readConfigFile(StringRef FileName,
for (Arg *A : *NewOptions)
A->claim();
+ // Filter out all -l and -Wl, options, put them into a separate list and
erase
+ // from the original list of configuration file options. These will be used
@@ -82,3 +82,29 @@
// CHECK-TWO-CONFIGS: -isysroot
// CHECK-TWO-CONFIGS-SAME: /opt/data
// CHECK-TWO-CONFIGS-SAME: -Wall
+
+//--- The linker input flags should be moved to the end of input list and
appear only when linking.
+// RUN: %clang --target=aarch64-unknown-linux-gnu --
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/117573
>From 65917ba219e26bd8547e8d1d9ff2c4ced1cdcad1 Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Mon, 25 Nov 2024 14:46:55 +
Subject: [PATCH] [clang][driver] Special care for -l and -Wl, flags in con
@@ -61,3 +61,29 @@
! CHECK-TWO-CONFIGS-NEXT: Configuration file:
{{.*}}Inputs{{.}}config2{{.}}config-4.cfg
! CHECK-TWO-CONFIGS: -ffp-contract=fast
! CHECK-TWO-CONFIGS: -O3
+
+!--- The linker input flags should be moved to the end of input list and
appear only when linking.
+!
@@ -82,3 +82,35 @@
// CHECK-TWO-CONFIGS: -isysroot
// CHECK-TWO-CONFIGS-SAME: /opt/data
// CHECK-TWO-CONFIGS-SAME: -Wall
+
+//--- The linker input flags should be moved to the end of input list and
appear only when linking.
+// RUN: %clang --target=aarch64-unknown-linux-gnu --
@@ -82,3 +82,35 @@
// CHECK-TWO-CONFIGS: -isysroot
// CHECK-TWO-CONFIGS-SAME: /opt/data
// CHECK-TWO-CONFIGS-SAME: -Wall
+
+//--- The linker input flags should be moved to the end of input list and
appear only when linking.
+// RUN: %clang --target=aarch64-unknown-linux-gnu --
pawosm-arm wrote:
> > I took a look and I think it may be too awkward to do, as we'd want to run
> > e.g. readelf afterwards. But an example is
> > `lld/test/ELF/as-needed-not-in-regular.s`. The idea being: if
> > `-Wl,--as-needed` is in the config file, do we correctly prune an
> > unnecessa
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/117573
>From b9dd5f9356d457ef1ad050083d650a1da21c4377 Mon Sep 17 00:00:00 2001
From: Pawel Osmialowski
Date: Mon, 25 Nov 2024 14:46:55 +
Subject: [PATCH] [clang][driver] Special care for -l and -Wl, flags in con
pawosm-arm wrote:
> @pawosm-arm See also
> https://wiki.gentoo.org/wiki/Project:Quality_Assurance/As-needed#Importance_of_linking_order.
> It is common for distributions to enable `-Wl,--as-needed` via patches to
> the compiler or configuration files.
Indeed, the distributions couldn't avoid
https://github.com/pawosm-arm updated
https://github.com/llvm/llvm-project/pull/116432
>From 12654b674725715422ba114e98bf76e40d557d16 Mon Sep 17 00:00:00 2001
From: Paul Osmialowski
Date: Tue, 3 Dec 2024 12:41:15 +
Subject: [PATCH] [clang][driver] When -fveclib=ArmPL flag is in use, always
@@ -102,3 +102,17 @@
/* Verify no warning when math-errno is re-enabled for a different veclib
(that does not imply -fno-math-errno). */
// RUN: %clang -### --target=aarch64-linux-gnu -fveclib=ArmPL -fmath-errno
-fveclib=LIBMVEC %s 2>&1 | FileCheck --check-prefix=CHECK-REPEAT-
@@ -490,6 +490,17 @@ void tools::AddLinkerInputs(const ToolChain &TC, const
InputInfoList &Inputs,
else
A.renderAsInput(Args, CmdArgs);
}
+ if (const Arg *A = Args.getLastArg(options::OPT_fveclib)) {
+if (A->getNumValues() == 1) {
pawosm-arm w
pawosm-arm wrote:
I've added these suggested changes. Also, I was advised by the libamath team
that the complicated relationship between libamath and libm needs more
elaborate handling, hence additional `-lm` before `-lamath` (see comment that
I've also added). Also I was advised to use `--as-
1 - 100 of 140 matches
Mail list logo
