Fix out-of-bounds access for 1x1 tridiagonal matrices and add shape checks by srinjoy933 · Pull Request #1162 · fortran-lang/stdlib

srinjoy933 · 2026-03-28T12:42:19Z

codecov · 2026-03-29T05:20:37Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 68.66%. Comparing base (72698fd) to head (d54c9cf).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1162   +/-   ##
=======================================
  Coverage   68.66%   68.66%           
=======================================
  Files         408      408           
  Lines       13619    13619           
  Branches     1537     1537           
=======================================
  Hits         9351     9351           
  Misses       4268     4268

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

srinjoy933 · 2026-03-29T12:03:40Z

HEY @jvdp1 @jalvesz what's your opinion about this issue? please suggest me if any changes are required .

Copilot

Pull request overview

Fixes edge cases and improves robustness in stdlib_specialmatrices tridiagonal support, addressing issue #1161 (1x1 dense conversion crash and missing dimension checks in arithmetic).

Changes:

Guard tridiagonal_to_dense_* against out-of-bounds access for n == 1.
Add dimension compatibility checks for tridiagonal add/sub operations.
Extend linalg special-matrices tests to cover the 1x1 dense conversion edge case and basic arithmetic.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
`src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp`	Prevents 1x1 out-of-bounds in dense conversion; adds dimension checks; refactors arithmetic construction paths.
`test/linalg/test_linalg_specialmatrices.fypp`	Adds tests for 1x1 dense conversion and tridiagonal arithmetic operators; updates suite naming/imports.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-29T12:31:05Z

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

+
+        if (A%n /= B%n) error stop "ValueError: tridiagonal matrices must have the same dimension to be added"
+
+        C = tridiagonal(A%dl + B%dl, A%dv + B%dv, A%du + B%du)


For consistency with the rest of the linalg/specialmatrices code, consider reporting the dimension-mismatch via linalg_state_type(..., LINALG_VALUE_ERROR, ...) + call linalg_error_handling(err0) (which is pure and already used in initialize_tridiagonal_pure_*) instead of a raw error stop with a Python-style ValueError: prefix. This will produce a standardized message including the component/location formatting used elsewhere in the library.

C = tridiagonal(A%dl + B%dl, A%dv + B%dv, A%du + B%du) : Same as before, this will trigger unnecessary copies. Please revert this change.

Regarding the raising the error, please look at linalg_state_type and use it for consistency with the rest of stdlib.

Copilot · 2026-03-29T12:31:05Z

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

+        if (A%n /= B%n) error stop "ValueError: tridiagonal matrices must have the same dimension to be added"
+
+        C = tridiagonal(A%dl + B%dl, A%dv + B%dv, A%du + B%du)
    end function


C = tridiagonal(A%dl + B%dl, ...) will typically create temporary arrays for each A%* +/- B%* expression and then copy again into C inside the constructor. If performance/memory is a concern for large n, prefer allocating/copying once (e.g., C = tridiagonal(A%dl, A%dv, A%du) after the size check) and then updating C%dl/dv/du in-place.

Copilot · 2026-03-29T12:31:05Z

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

+
+        if (A%n /= B%n) error stop "ValueError: tridiagonal matrices must have the same dimension to be subtracted"
+
+        C = tridiagonal(A%dl - B%dl, A%dv - B%dv, A%du - B%du)


Same as for addition: consider using linalg_state_type + linalg_error_handling for the dimension mismatch so error reporting is consistent across this submodule, rather than a standalone error stop string.

Same as before. Copilot is correct again.

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

test/linalg/test_linalg_specialmatrices.fypp

loiseaujc

Nice catch ! It never occurred to me that 1 x 1 matrices might be a particular edge case of tridiagonal matrices. It good overall, but there are a few bits and pieces that you need to revert to the original state for performance-sake, and look at linalg_state_type and linalg_error_handling to handle the errors is an stdlib-consistent way.

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

loiseaujc · 2026-03-30T05:38:46Z

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

+
+        if (A%n /= B%n) error stop "ValueError: tridiagonal matrices must have the same dimension to be added"
+
+        C = tridiagonal(A%dl + B%dl, A%dv + B%dv, A%du + B%du)


C = tridiagonal(A%dl + B%dl, A%dv + B%dv, A%du + B%du) : Same as before, this will trigger unnecessary copies. Please revert this change.

Regarding the raising the error, please look at linalg_state_type and use it for consistency with the rest of stdlib.

loiseaujc · 2026-03-30T05:39:21Z

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

+
+        if (A%n /= B%n) error stop "ValueError: tridiagonal matrices must have the same dimension to be subtracted"
+
+        C = tridiagonal(A%dl - B%dl, A%dv - B%dv, A%du - B%du)


Same as before. Copilot is correct again.

srinjoy933 · 2026-03-30T15:58:47Z

hey @loiseaujc I have made the changes that you have suggested , please have a look on this pr and suggest me if any further changes are required .Thank you!

jvdp1

thank you for this PR and fix. Please find here some minor suggestions

test/linalg/test_linalg_specialmatrices.fypp

jvdp1 · 2026-03-30T17:08:52Z

test/linalg/test_linalg_specialmatrices.fypp

+            allocate(dl(0), dv(1), du(0))
+            dv(1) = 5.0_wp


dl, dv and du could be defined as parameter

jvdp1 · 2026-03-30T17:09:29Z

test/linalg/test_linalg_specialmatrices.fypp

+            allocate(dl1(n-1), dv1(n), du1(n-1))
+            allocate(dl2(n-1), dv2(n), du2(n-1))
+
+            dl1 = 1.0_wp ; dv1 = 2.0_wp ; du1 = 3.0_wp
+            dl2 = 4.0_wp ; dv2 = 5.0_wp ; du2 = 6.0_wp


These variables could be defined as parameter

srinjoy933 · 2026-04-01T18:31:02Z

hey @loiseaujc @jvdp1 I have made the changes that you have suggested please have a look at this pr,if it requires any more changes please suggest me

jvdp1

Thank you @srinjoy933 for this fix. I think that it is ready to be merged.

test/linalg/test_linalg_specialmatrices.fypp

Mahmood-Sinan

@srinjoy933 Here are my views.

Mahmood-Sinan · 2026-04-03T16:25:20Z

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

+        if (n == 1) then
+            B(1, 1) = A%dv(1)
+        else
+            B(1, 1) = A%dv(1) ; B(1, 2) = A%du(1)


Suggested change

B(1, 1) = A%dv(1) ; B(1, 2) = A%du(1)

B(1, 1) = A%dv(1)

B(1, 2) = A%du(1)

Mahmood-Sinan · 2026-04-03T16:25:33Z

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

+                B(i, i) = A%dv(i)
+                B(i, i+1) = A%du(i)
+            enddo
+            B(n, n-1) = A%dl(n-1) ; B(n, n) = A%dv(n)


Suggested change

B(n, n-1) = A%dl(n-1) ; B(n, n) = A%dv(n)

B(n, n-1) = A%dl(n-1)

B(n, n) = A%dv(n)

Mahmood-Sinan · 2026-04-03T16:25:52Z

test/linalg/test_linalg_specialmatrices.fypp

+    subroutine test_tridiagonal_1x1(error)
+        !> Test 1x1 matrix edge case for dense conversion
+        type(error_type), allocatable, intent(out) :: error
+        #:for k1, t1, s1 in (KINDS_TYPES)


Suggested change

#:for k1, t1, s1 in (KINDS_TYPES)

#:for k1, t1, s1 in KINDS_TYPES

Please fetch/pull once, because I've made some changes in the module. One of them was removing the brackets around KINDS_TYPES. For consistency, it would be better to remove the brackets.

Mahmood-Sinan · 2026-04-03T16:26:00Z

test/linalg/test_linalg_specialmatrices.fypp

+            type(tridiagonal_${s1}$_type) :: A
+            ${t1}$, allocatable :: Amat(:,:)
+
+            ! Uses parameter with standard F2003 empty array syntax to satisfy maintainer


What do you mean by 'to satisfy maintainer' ?

Actually I have a habit to write comments to understand what changes I have made now , when I have already done multiple commits for a particular file.I forgot to erase off those comments in this case .Nothing else. Really thank you for pointing out this mistake .

Mahmood-Sinan · 2026-04-03T16:26:09Z

test/linalg/test_linalg_specialmatrices.fypp

+    subroutine test_tridiagonal_arithmetic(error)
+        !> Test arithmetic operations and optimization
+        type(error_type), allocatable, intent(out) :: error
+        #:for k1, t1, s1 in (KINDS_TYPES)


Suggested change

#:for k1, t1, s1 in (KINDS_TYPES)

#:for k1, t1, s1 in KINDS_TYPES

Mahmood-Sinan · 2026-04-03T16:26:14Z

test/linalg/test_linalg_specialmatrices.fypp

+            integer, parameter :: n = 3
+            type(tridiagonal_${s1}$_type) :: A, B, C
+
+            ! Uses explicit parameter lists to satisfy maintainer and avoid 'i' compiler bug


Again, 'to satisfy maintainer' ?

same cause that I have previously mentioned

Mahmood-Sinan · 2026-04-03T16:29:54Z

test/linalg/test_linalg_specialmatrices.fypp

+            ! Check if it compiled and converted properly without segfaulting
+            call check(error, size(Amat, 1) == 1, .true.)


How can compile error be detected by this code ? If there is a compile error, that would be detected while compiling/building, not while running.

Thank you for pointing out this. You're completely right. I've adjusted the wording to simply state: ! Check if the 1x1 matrix converted properly at runtime without segfaulting

srinjoy933 · 2026-04-04T13:54:43Z

hey @Mahmood-Sinan Thanks for your valuable suggestions. I have made the changes that you have suggested ,please have a look at this pr

Mahmood-Sinan

@srinjoy933 Thanks for the rapid changes. A couple more changes and i think it will be good. One more suggestion is that, since the 1x1 test and arithmetic test for tridiagonal uses parameter variables, would you mind doing the same for the 1x1 test and arithmetic test for sym tridiagonal ? It would be better for consistency.

Mahmood-Sinan · 2026-04-05T05:51:23Z

src/specialmatrices/stdlib_specialmatrices_tridiagonal.fypp

        type(tridiagonal_${s1}$_type) :: A
        !! Corresponding tridiagonal matrix.
-
+        


Please remove the trailing whitespace.

Mahmood-Sinan · 2026-04-05T05:53:14Z

test/linalg/test_linalg_specialmatrices.fypp

                "spmv(fail): y = alpha*A*x + beta*y, alpha: "//to_string(alpha)//", beta: "//to_string(beta))
            if (allocated(error)) return

-            ! Test y = alpha * A.T @ x beta * y for random values of alpha and beta


Thanks for pointing it out.

Mahmood-Sinan · 2026-04-05T05:55:46Z

test/linalg/test_linalg_specialmatrices.fypp


    testsuites = [ &
-        new_testsuite("specialmatrices", collect_suite) &
+        new_testsuite("special matrices", collect_suite) &


Suggested change

new_testsuite("special matrices", collect_suite) &

new_testsuite("special_matrices", collect_suite) &

srinjoy933 · 2026-04-05T10:10:37Z

Thanks for the valuable suggestions, @Mahmood-Sinan! I'd really like to implement the same thing for symmetric tridiagonal matrices, but I think it would be best to handle that in a separate PR. If everything looks good to go here, we can merge this one and I'll tackle the symmetric tridiagonal updates next. Let me know what you think!

srinjoy933 · 2026-04-05T10:19:50Z

@Mahmood-Sinan I have made the changes that you have suggested. Please have a look at this pr ,and kindly suggest if any more changes are required. Thank You!

Mahmood-Sinan

Thanks you for the PR. LGTM.

Fix tridiagonal 1x1 segfault and optimize arithmetic operations

9e200ee

srinjoy933 marked this pull request as draft March 28, 2026 12:44

updated 2

9afb581

srinjoy933 marked this pull request as ready for review March 29, 2026 04:34

jalvesz requested review from Copilot and loiseaujc March 29, 2026 12:27

Copilot started reviewing on behalf of jalvesz March 29, 2026 12:28 View session

Copilot AI reviewed Mar 29, 2026

View reviewed changes

loiseaujc requested changes Mar 30, 2026

View reviewed changes

loiseaujc mentioned this pull request Mar 30, 2026

Add symmetric tridiagonal matrices module #1091

Merged

1 task

update

f5e185b

srinjoy933 requested a review from loiseaujc March 30, 2026 15:59

jvdp1 reviewed Mar 30, 2026

View reviewed changes

new

8888f66

srinjoy933 requested a review from jvdp1 April 1, 2026 16:26

srinjoy933 added 3 commits April 1, 2026 22:08

changed

52971c3

new2

e8f11ed

changed

e2dac45

jvdp1 approved these changes Apr 1, 2026

View reviewed changes

test/linalg/test_linalg_specialmatrices.fypp Show resolved Hide resolved

loiseaujc approved these changes Apr 2, 2026

View reviewed changes

Mahmood-Sinan reviewed Apr 3, 2026

View reviewed changes

srinjoy933 added 3 commits April 4, 2026 19:10

updated

f040d34

changed

0cc6207

Merge branch 'master' into fix-tridiagonal-edge-cases

42054c9

srinjoy933 requested a review from Mahmood-Sinan April 4, 2026 13:53

srinjoy933 added 2 commits April 4, 2026 19:54

done

9816d8a

done

ecc8ecf

srinjoy933 marked this pull request as draft April 4, 2026 14:42

changed

6802a73

srinjoy933 marked this pull request as ready for review April 4, 2026 17:03

Mahmood-Sinan reviewed Apr 5, 2026

View reviewed changes

updated

d54c9cf

srinjoy933 requested a review from Mahmood-Sinan April 5, 2026 10:18

Mahmood-Sinan approved these changes Apr 5, 2026

View reviewed changes


		if (A%n /= B%n) error stop "ValueError: tridiagonal matrices must have the same dimension to be added"

		C = tridiagonal(A%dl + B%dl, A%dv + B%dv, A%du + B%du)


		if (A%n /= B%n) error stop "ValueError: tridiagonal matrices must have the same dimension to be subtracted"

		C = tridiagonal(A%dl - B%dl, A%dv - B%dv, A%du - B%du)

	B(1, 1) = A%dv(1) ; B(1, 2) = A%du(1)
	B(1, 1) = A%dv(1)
	B(1, 2) = A%du(1)

	B(n, n-1) = A%dl(n-1) ; B(n, n) = A%dv(n)
	B(n, n-1) = A%dl(n-1)
	B(n, n) = A%dv(n)

	#:for k1, t1, s1 in (KINDS_TYPES)
	#:for k1, t1, s1 in KINDS_TYPES

		! Check if it compiled and converted properly without segfaulting
		call check(error, size(Amat, 1) == 1, .true.)

		type(tridiagonal_${s1}$_type) :: A
		!! Corresponding tridiagonal matrix.

	new_testsuite("special matrices", collect_suite) &
	new_testsuite("special_matrices", collect_suite) &

Conversation

srinjoy933 commented Mar 28, 2026

Uh oh!

codecov bot commented Mar 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

srinjoy933 commented Mar 29, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 29, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 29, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 29, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

loiseaujc left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

srinjoy933 commented Mar 30, 2026

Uh oh!

jvdp1 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

srinjoy933 commented Apr 1, 2026

Uh oh!

jvdp1 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Mahmood-Sinan 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!

Choose a reason for hiding this comment

Uh oh!

srinjoy933 Apr 4, 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!

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!

srinjoy933 commented Apr 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

codecov bot commented Mar 29, 2026 •

edited

Loading

srinjoy933 Apr 4, 2026 •

edited

Loading

srinjoy933 commented Apr 4, 2026 •

edited

Loading

srinjoy933 commented Apr 5, 2026 •

edited

Loading

srinjoy933 commented Apr 5, 2026 •

edited

Loading