`utils.c`: include `<signal.h>` for `siginfo_t` #7517

aitap · 2025-12-25T19:48:53Z

POSIX says:

The <signal.h> header shall define the siginfo_t type as a structure

So <sys/wait.h> is not enough to see the definition (not just a forward declaration) of siginfo_t.

Tested manually on FreeBSD:

Details

$ R CMD check data.table_1.18.99.tar.gz --no-manual

* using R version 4.5.1 (2025-06-13)                                                   
* using platform: amd64-portbld-freebsd15.0                                            
* R was compiled by                                                                                                                                                            
    FreeBSD clang version 19.1.7 (https://github.com/llvm/llvm-project.git llvmorg-19.1.7-0-gcd708029e0b2)
    GNU Fortran (FreeBSD Ports Collection) 13.3.0                                      
* running under: FreeBSD freebsd 15.0-RELEASE FreeBSD 15.0-RELEASE releng/15.0-n280995-7aedc8de6446 GENERIC amd64
* checking whether package ‘data.table’ can be installed ... OK                        
* used C compiler: ‘FreeBSD clang version 19.1.7 (https://github.com/llvm/llvm-project.git llvmorg-19.1.7-0-gcd708029e0b2)’                                                    
* checking compiled code ... NOTE                                                      
File ‘data.table/libs/data_table.so’:   
  Found non-API calls to R: ‘IS_GROWABLE’, ‘LEVELS’, ‘SETLENGTH’,
    ‘SET_GROWABLE_BIT’, ‘SET_TRUELENGTH’, ‘TRUELENGTH’, ‘XTRUELENGTH’
Status: 1 NOTE

...and OpenBSD:

Details

$ LANG=C.UTF-8 R CMD check data.table_1.18.99.tar.gz --no-manual

* R was compiled by                                                                    
    egcc (GCC) 8.4.0                                                                   
    GNU Fortran (GCC) 8.4.0                                                            
* running under: OpenBSD openbsd.my.domain 7.8 GENERIC#54 amd64                        

* checking whether package ‘data.table’ can be installed ... OK
* used C compiler: ‘OpenBSD clang version 19.1.7’

Status: OK

Also tested on GNU/Linux with PKG_CFLAGS=--std=c99.

@jszhao, could you please pull from this branch (posix_siginfo_t) and run R CMD build, then install the resulting tarball?

POSIX says: > The <signal.h> header shall define the siginfo_t type as a structure So <sys/wait.h> is not enough to see the definition (not just a forward declaration) of siginfo_t.

codecov · 2025-12-25T19:55:08Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.02%. Comparing base (11a9435) to head (3cc6658).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #7517   +/-   ##
=======================================
  Coverage   99.02%   99.02%           
=======================================
  Files          87       87           
  Lines       16803    16803           
=======================================
  Hits        16640    16640           
  Misses        163      163

☔ 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.

NEWS.md

github-actions · 2025-12-25T20:10:57Z

HEAD=posix_siginfo_t slower P<0.001 for DT[by,verbose=TRUE] improved in #6296

Generated via commit 3cc6658

Download link for the artifact containing the test results: ↓ atime-results.zip

Task	Duration
R setup and installing dependencies	3 minutes and 3 seconds
Installing different package versions	44 seconds
Running and plotting the test cases	4 minutes and 29 seconds

src/utils.c

jszhao · 2025-12-26T05:59:08Z

POSIX says:

The <signal.h> header shall define the siginfo_t type as a structure

So <sys/wait.h> is not enough to see the definition (not just a forward declaration) of siginfo_t.

Fixes: #7516

Tested manually on FreeBSD:
Details

...and OpenBSD:
Details

Also tested on GNU/Linux with PKG_CFLAGS=--std=c99.

@jszhao, could you please pull from this branch (posix_siginfo_t) and run R CMD build, then install the resulting tarball?

I have used R CMD INSTALL ... to install this branch (in fact, I am not familiar with github, so I just downloaded the branch with zip file, unzip, and tar it to tarball). It works. Thanks a lot for your effect to fix it for FreeBSD. Thanks again.

MichaelChirico

Great!

MichaelChirico · 2026-01-01T22:52:25Z

cc @aitap for extra eyes on a5a2644 -- should we add the same to the other jobs with -D_FORTIFY_SOURCE=2? just doing so in the strict job for now.

aitap · 2026-01-03T19:31:19Z

Ironically, setting -D_POSIX_C_SOURCE=200809L makes gettimeofday invisible in FreeBSD 15 (indeed, POSIX.1-2008 marks it as obsolete, recommending clock_gettime() instead):

user@freebsd:~ $ echo '#include <sys/time.h>' | cpp | grep gettimeofday                                   
int gettimeofday(struct timeval *, struct timezone *);
user@freebsd:~ $ echo '#include <sys/time.h>' | cpp -D_POSIX_C_SOURCE=200809L | grep gettimeofday || echo not found
not found

We use gettimeofday in wallclock. (The code that wants to use clock_gettime() deprives itself of the opportunity by testing for #ifdef CLOCK_REALTIME before a header file can #define it. A fix to that will need to be careful not to reintroduce #5374.)

Unlike _POSIX_C_SOURCE, which controls symbol visibility, _FORTIFY_SOURCE rewrites some function calls to perform stricter checks. Is the idea to make strict checks use #define _POSIX_C_SOURCE 200809L because it will hide non-standard or obsolete symbols?

MichaelChirico · 2026-01-03T19:41:41Z

I'm just looking for a way to fortify our checks such that we would have caught this issue before release

aitap · 2026-01-03T20:12:16Z

Then the strict checks must use --std=c99 to hide even more symbols, so that data.table could request some of them back using #define _POSIX_C_SOURCE=200809L:
https://gitlab.com/Rdatatable/data.table/-/commit/f02dcb783903727cbc043620f764b9d7f25c7af9/pipelines?ref=stricter_c99

~~(Right now the only failure is the -vanilla test, due to the C locale.)~~

Should I push f02dcb7 to this branch?

Drat. R itself uses configure-time tests for C23 features in R_ext/Rboolean.h, so when we later request C99 compatibility, this causes annoying warnings (because the configure-time tests aren't updated).

ben-schwen · 2026-01-12T13:18:36Z

Since we need to patch 1.18.2 soon, I would be in favor to merge this and check afterwards if new errors pop up (I know this is not ideal)

.gitlab-ci.yml

* utils.c: include <signal.h> for siginfo_t POSIX says: > The <signal.h> header shall define the siginfo_t type as a structure So <sys/wait.h> is not enough to see the definition (not just a forward declaration) of siginfo_t. * NEWS entry * Amend NEWS * more robustly define _POSIX_C_SOURCE (h/t Hugh) * tidy up NEWS * -D_POSIX_C_SOURCE=200809L in gitlab CI job for regression test * revert gitlab-ci change --------- Co-authored-by: Michael Chirico <[email protected]> Co-authored-by: Michael Chirico <[email protected]>

MichaelChirico · 2026-01-12T17:52:14Z

Also cherry-picked to the 1.18.2 patch: 20e25d1

aitap · 2026-01-12T19:43:01Z

NEWS.md


 6. `rowwiseDT()` now provides a helpful error message when a complex object that is not a list (e.g., a function) is provided as a cell value, instructing the user to wrap it in `list()`, [#7219](https://github.com/Rdatatable/data.table/issues/7219). Thanks @kylebutts for the report and @venom1204 for the fix.

+7. Fixed compilation failure like "error: unknown type name 'siginfo_t'" in v1.18.0 in some strict environments, e.g., FreeBSD, where the header file declaring the POSIX function `waitid` does not transitively include the header file defining the `siginfo_t` type, [#7516](https://github.com/rdatatable/data.table/issues/7516). Thanks to @jszhao for the report and @aitap for the fix.


Very minor: the underscore in "error: unknown type name 'siginfo_t'" is an unpaired emphasis mark. This doesn't confuse Pandoc or commonmark, but makes the Vim syntax highlighter unhappy. Some variation on "error: unknown type name 'siginfo\_t'" or "error: unknown type name `siginfo_t`" would be slightly better.

I would prefer `error: unknown type name 'siginfo_t'` to keep it human-readable too

aitap added 2 commits December 25, 2025 21:54

utils.c: include <signal.h> for siginfo_t

2da9fd8

POSIX says: > The <signal.h> header shall define the siginfo_t type as a structure So <sys/wait.h> is not enough to see the definition (not just a forward declaration) of siginfo_t.

NEWS entry

513b017

aitap requested review from HughParsonage and MichaelChirico as code owners December 25, 2025 19:48

MichaelChirico approved these changes Dec 25, 2025

View reviewed changes

MichaelChirico reviewed Dec 25, 2025

View reviewed changes

NEWS.md Outdated Show resolved Hide resolved

Amend NEWS

baaf8ca

HughParsonage reviewed Dec 25, 2025

View reviewed changes

src/utils.c Show resolved Hide resolved

MichaelChirico added 2 commits December 31, 2025 00:47

more robustly define _POSIX_C_SOURCE (h/t Hugh)

db69eae

tidy up NEWS

327e2a5

MichaelChirico approved these changes Dec 31, 2025

View reviewed changes

MichaelChirico added 2 commits December 31, 2025 00:50

Merge branch 'master' into posix_siginfo_t

970c0bf

-D_POSIX_C_SOURCE=200809L in gitlab CI job for regression test

a5a2644

MichaelChirico requested review from ben-schwen and jangorecki as code owners January 1, 2026 22:51

Merge branch 'master' into posix_siginfo_t

08771f1

Merge branch 'master' into posix_siginfo_t

7e3fbc9

MichaelChirico reviewed Jan 12, 2026

View reviewed changes

.gitlab-ci.yml Outdated Show resolved Hide resolved

revert gitlab-ci change

3cc6658

MichaelChirico merged commit a325db9 into master Jan 12, 2026
13 checks passed

MichaelChirico deleted the posix_siginfo_t branch January 12, 2026 17:50

aitap commented Jan 12, 2026

View reviewed changes


		6. `rowwiseDT()` now provides a helpful error message when a complex object that is not a list (e.g., a function) is provided as a cell value, instructing the user to wrap it in `list()`, [#7219](https://github.com/Rdatatable/data.table/issues/7219). Thanks @kylebutts for the report and @venom1204 for the fix.

		7. Fixed compilation failure like "error: unknown type name 'siginfo_t'" in v1.18.0 in some strict environments, e.g., FreeBSD, where the header file declaring the POSIX function `waitid` does not transitively include the header file defining the `siginfo_t` type, [#7516](https://github.com/rdatatable/data.table/issues/7516). Thanks to @jszhao for the report and @aitap for the fix.

utils.c: include <signal.h> for siginfo_t #7517

utils.c: include <signal.h> for siginfo_t #7517

Conversation

aitap commented Dec 25, 2025

Uh oh!

codecov bot commented Dec 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

github-actions bot commented Dec 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

jszhao commented Dec 26, 2025

Uh oh!

MichaelChirico left a comment

Choose a reason for hiding this comment

Uh oh!

MichaelChirico commented Jan 1, 2026

Uh oh!

aitap commented Jan 3, 2026

Uh oh!

MichaelChirico commented Jan 3, 2026

Uh oh!

aitap commented Jan 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ben-schwen commented Jan 12, 2026

Uh oh!

Uh oh!

Uh oh!

MichaelChirico commented Jan 12, 2026

Uh oh!

aitap Jan 12, 2026

Choose a reason for hiding this comment

Uh oh!

MichaelChirico Jan 12, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

`utils.c`: include `<signal.h>` for `siginfo_t` #7517

`utils.c`: include `<signal.h>` for `siginfo_t` #7517

codecov bot commented Dec 25, 2025 •

edited

Loading

github-actions bot commented Dec 25, 2025 •

edited

Loading

aitap commented Jan 3, 2026 •

edited

Loading