test: add tests and docs for section address linker options by sai18022001 · Pull Request #1145 · qualcomm/eld

sai18022001 · 2026-05-07T10:50:00Z

Add missing SectionStart.test for Hexagon target
Add Common tests for -Ttext, -Tdata, -Tbss
Add Common test for -Ttext-segment
Add x86_64 tests for -Trodata-segment and -Tldata-segment
Update userguide layout.rst with descriptions for all options

Modifications to support this fix:
Add -Trodata-segment and -Tldata-segment option definitions and implementation
Implement -Ttext-segment option (required for tests to pass)

Note: When both -Ttext-segment and a linker script address assignment are
specified, the linker script takes precedence. This matches GNU ld behavior
and has been confirmed by the maintainer.

- Add tests for --section-start, -Ttext, -Tdata, -Tbss, -Ttext-segment - Add -Trodata-segment and -Tldata-segment support and tests - Update userguide with option descriptions Resolves qualcomm#1099 Signed-off-by: Sai Sanjay Chikne <saichikne180201@gmail.com>

quic-seaswara · 2026-05-07T16:21:35Z


 uint64_t GNULDBackend::getImageBase(bool HasInterp, bool LoadEHdr) const {
+  if (auto TextSegment = config().options().textSegment())
+    return *TextSegment;


Does -Ttext-segment override linker script assignment ?

I tested this, when both -Ttext-segment and a linker script address assignment are specified, the linker script takes precedence. Is this the expected behavior, or should -Ttext-segment always override the linker script ?

Pinging on this - wanted to confirm the expected behavior before making any further changes to GNULDBackend.cpp.
Based on GNU ld behavior, the linker script takes precedence over -Ttext-segment when an explicit address is assigned in the script. The current implementation follows this behavior. Please let me know if a different behavior is expected for ELD.

Thanks for confirming. Please go ahead.

parth-07 · 2026-05-08T10:04:11Z


  void setImageBase(uint64_t Value) { ImageBase = Value; }
+
+  const std::optional<uint64_t> &textSegment() const { return TextSegment; }


Why is the PR for adding tests and docs modifying functionality?

Good point.
Tests wouldn't pass without -Text-segment, -Trodata-segment and -Tldata-segment as it were not implemented previously
should i split them into 2 PR - one for tests/docs and one for implemetation ?
or you want to have a look at those implementations ?

I think what @parth-07 is mentioning is the commit message says adding tests but now we are implementing support. The issue needs to be properly documented and also the commit message.

I'll update the commit message. Thanks for clarifying.

…options - Add tests for --section-start, -Ttext, -Tdata, -Tbss, -Ttext-segment - Add -Trodata-segment and -Tldata-segment support and tests - Implement -Ttext-segment, -Trodata-segment, -Tldata-segment options - Update userguide with option descriptions Resolves qualcomm#1099 Signed-off-by: Sai Sanjay Chikne <saichikne180201@gmail.com>

sai18022001 · 2026-05-13T09:34:23Z

@quic-seaswara @parth-07
All review comments have been addressed. Requesting a re-review when you get a chance.

Signed-off-by: Sai Sanjay Chikne <saichikne180201@gmail.com>

quic-seaswara · 2026-05-13T20:45:18Z


+--section-start=sectionname=org
+    Assigns the absolute address org to the named output section.
+    You may use this option multiple times to place multiple sections at specific addresses.


What happens if you specify the --section-start with the same section name multiple times ? Can we add that to the documentation ?

quic-seaswara · 2026-05-13T20:45:29Z

+    You may use this option multiple times to place multiple sections at specific addresses.
+
+-Ttext=org
+    Same as ``--section-start`` with ``.text`` as the section name.


What if there is no .text section ?

quic-seaswara · 2026-05-13T20:45:42Z

+    Same as ``--section-start`` with ``.text`` as the section name.
+
+-Tdata=org
+    Same as ``--section-start`` with ``.data`` as the section name.


Same as previous, what if there is no data section ?

quic-seaswara · 2026-05-13T20:46:26Z

+  if (llvm::opt::Arg *arg = Args.getLastArg(T::Trodata_segment)) {
+    uint64_t addr = 0;
+    if (!llvm::StringRef(arg->getValue()).getAsInteger(0, addr))
+      Config.options().setRodataSegment(addr);


Is this handled in program layout ?

It is currently parsed and stored but not yet handled in program layout. Would you like me to implement it in this PR or open a separate issue for it?

quic-seaswara · 2026-05-13T20:46:36Z

+  if (llvm::opt::Arg *arg = Args.getLastArg(T::Tldata_segment)) {
+    uint64_t addr = 0;
+    if (!llvm::StringRef(arg->getValue()).getAsInteger(0, addr))
+      Config.options().setLdataSegment(addr);


Is this handled in program layout ?

It is currently parsed and stored but not yet handled in program layout. Would you like me to implement it in this PR or open a separate issue for it?

quic-seaswara · 2026-05-13T20:47:13Z

+  if (llvm::opt::Arg *arg = Args.getLastArg(T::Ttext_segment)) {
+    uint64_t addr = 0;
+    if (!llvm::StringRef(arg->getValue()).getAsInteger(0, addr))
+      Config.options().setTextSegment(addr);


It might be useful to give a warning with linker scripts if the user used a linker script and users using this option.

quic-seaswara · 2026-05-13T20:47:59Z

+  if (llvm::opt::Arg *arg = Args.getLastArg(T::Ttext_segment)) {
+    uint64_t addr = 0;
+    if (!llvm::StringRef(arg->getValue()).getAsInteger(0, addr))
+      Config.options().setTextSegment(addr);


Also add a diagnostic if the value is not represented as a integer or does not fit the target address space.

For example using a 64bit address on a 32bit image.

quic-seaswara · 2026-05-13T20:48:31Z

      Group<grp_scriptopts>;
+defm Trodata_segment
+    : dashEqWithOpt<"Trodata-segment", "Trodata-segment", "Trodata_segment",
+                    "Specify an address for the .rodata-segment segment">,


Can you reword this ? Is there anything like .rodata-segment ?

quic-seaswara · 2026-05-13T20:48:55Z

+      Group<grp_scriptopts>;
+defm Tldata_segment
+    : dashEqWithOpt<"Tldata-segment", "Tldata-segment", "Tldata_segment",
+                    "Specify an address for the .ldata-segment segment">,


What is .ldata-segment ? Can you simplify this and reword ?

quic-seaswara · 2026-05-13T20:49:57Z


  void setImageBase(uint64_t Value) { ImageBase = Value; }
+
+  const std::optional<uint64_t> &textSegment() const { return TextSegment; }


Can we convert these functions to use verbs, get/set.

I recommend using getTextSegmentAddress(), it would be very clear.

Also update the text map file so that an user inspecting the text map file can know what is going on.

- Rename textSegment/rodataSegment/ldataSegment to getTextSegmentAddress/getRodataSegmentAddress/getLdataSegmentAddress - Fix -Trodata-segment and -Tldata-segment option descriptions - Update --section-start, -Ttext, -Tdata, -Tbss documentation Signed-off-by: Sai Sanjay Chikne <saichikne180201@gmail.com>

parth-07

@sai18022001 Can you please move the tests to test/Common subdirectory?

Additionally, can you please update the PR description and the commit message to refect that this PR is adding the functionality instead of just adding test for the functionality.

parth-07 · 2026-05-26T13:23:16Z

+  // -Ttext-segment=value
+  if (llvm::opt::Arg *arg = Args.getLastArg(T::Ttext_segment)) {
+    uint64_t addr = 0;
+    if (!llvm::StringRef(arg->getValue()).getAsInteger(0, addr))


We should report an error if the argument is incorrect for the new options added.

parth-07 · 2026-05-26T13:26:28Z

 }

 uint64_t GNULDBackend::getImageBase(bool HasInterp, bool LoadEHdr) const {
+  if (auto TextSegment = config().options().getTextSegmentAddress())


Can you please check what is the precedence between --image-base and -Ttext-segment in GNU ld?

parth-07 · 2026-05-26T13:28:43Z

Why is this test hexagon specific?

sai18022001 force-pushed the test/section-start-options branch from 9ac24b2 to 3eba23f Compare May 7, 2026 10:52

sai18022001 mentioned this pull request May 7, 2026

Write tests to test and evaluate the following linker options #1099

Open

quic-seaswara reviewed May 7, 2026

View reviewed changes

parth-07 reviewed May 8, 2026

View reviewed changes

sai18022001 requested review from parth-07 and quic-seaswara May 10, 2026 08:43

docs: fix --section-start description wording in userguide

ed54426

Signed-off-by: Sai Sanjay Chikne <saichikne180201@gmail.com>

sai18022001 force-pushed the test/section-start-options branch from 40cf58c to ed54426 Compare May 13, 2026 09:45

quic-seaswara requested changes May 13, 2026

View reviewed changes

sai18022001 requested a review from quic-seaswara May 14, 2026 10:57

parth-07 reviewed May 26, 2026

View reviewed changes


		void setImageBase(uint64_t Value) { ImageBase = Value; }

		const std::optional<uint64_t> &textSegment() const { return TextSegment; }

Conversation

sai18022001 commented May 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sai18022001 May 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sai18022001 May 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sai18022001 commented May 13, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

parth-07 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

sai18022001 commented May 7, 2026 •

edited

Loading

sai18022001 May 9, 2026 •

edited

Loading

sai18022001 May 8, 2026 •

edited

Loading