src: only block SIGUSR1 when HAVE_INSPECTOR #12266

danbev · 2017-04-07T06:54:58Z

I'm currently seeing a timeout error for test-signal-handler.js on
macosx when using the following configuration:

./configure --debug --without-ssl && make -j8 test

--without-ssl implies that there will be no inspector but the signal
SIGUSR1 is blocked in PlatformInit just the same. But in this case
never unblocked which is causing the signal to never be delivered to
the handlers in test-signal-handler.js and it loops until it times out.

Not sure if this is the best way of fixing this but hopefully more eyes
on this will help.

Checklist

make -j4 test (UNIX), or vcbuild test (Windows) passes
commit message follows commit guidelines

Affected core subsystem(s)

src

danbev · 2017-04-07T06:55:44Z

CI: https://ci.nodejs.org/job/node-test-pull-request/7261/

bnoordhuis

One more for nodejs/build#419.

bnoordhuis · 2017-04-07T08:14:15Z

src/node.cc

  sigset_t sigmask;
  sigemptyset(&sigmask);
  sigaddset(&sigmask, SIGUSR1);
  const int err = pthread_sigmask(SIG_SETMASK, &sigmask, nullptr);
+  CHECK_EQ(err, 0);
+#endif


Can you write this as #endif // HAVE_INSPECTOR? (Note the two blanks before the comment.)

You can't really do the CHECK_EQ at this point because fds 0-2 might not be pointing to valid file descriptions yet.

danbev · 2017-04-07T08:24:16Z

Can you write this as #endif // HAVE_INSPECTOR? (Note the two blanks before the comment.)

Yep, I'll fix that.

You can't really do the CHECK_EQ at this point because fds 0-2 might not be pointing to valid file descriptions yet.

Sorry I'm missing something here :( CHECK_EQ is checking the return value from pthread_sigmask and err is const. How do the file descriptors effect this?

bnoordhuis · 2017-04-07T08:35:23Z

Check the for (int fd = STDIN_FILENO; fd <= STDERR_FILENO; fd += 1) loop right below it.

danbev · 2017-04-07T08:47:58Z

Check the for (int fd = STDIN_FILENO; fd <= STDERR_FILENO; fd += 1) loop right below it.

Oh! Let's not speak of this again :)

danbev · 2017-04-07T09:46:39Z

CI: https://ci.nodejs.org/job/node-test-pull-request/7263/

bnoordhuis · 2017-04-07T09:49:03Z

src/node.cc

  CHECK_EQ(err, 0);
+#endif  // HAVE_INSPECTOR


I don't want to make this a torment for you but the reason SIGUSR1 is blocked first thing is to make the 'kill window' as small as possible. Some tools start node and immediately send a SIGUSR1 - if the signal isn't masked, it kills the process.

Ah I see, that makes sense. I was trying to avoid the additional HAVE_INSTRUCTOR guard but that is no problem.

I don't want to make this a torment

No torment at all, I appreciate all the reviews and feedback!

I'm currently seeing a timeout error for test-signal-handler.js on macosx when using the following configuration: ./configure --debug --without-ssl && make -j8 test --without-ssl implies that there will be no inspector but the signal SIGUSR1 is blocked in PlatformInit just the same. But in this case never unblocked which is causing the signal to never be delivered to the handlers in test-signal-handler.js and it loops until it times out. Not sure if this is the best way of fixing this but hopefully more eyes on this will help.

danbev · 2017-04-09T17:05:13Z

CI: https://ci.nodejs.org/job/node-test-pull-request/7293/

I'm currently seeing a timeout error for test-signal-handler.js on macosx when using the following configuration: ./configure --debug --without-ssl && make -j8 test --without-ssl implies that there will be no inspector but the signal SIGUSR1 is blocked in PlatformInit just the same. But in this case never unblocked which is causing the signal to never be delivered to the handlers in test-signal-handler.js and it loops until it times out. Not sure if this is the best way of fixing this but hopefully more eyes on this will help. PR-URL: nodejs#12266 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl> Reviewed-By: Anna Henningsen <anna@addaleax.net>

danbev · 2017-04-10T05:23:12Z

Landed in 65a6e05

italoacasas · 2017-04-10T22:16:55Z

cc @danbev

I'm currently seeing a timeout error for test-signal-handler.js on macosx when using the following configuration: ./configure --debug --without-ssl && make -j8 test --without-ssl implies that there will be no inspector but the signal SIGUSR1 is blocked in PlatformInit just the same. But in this case never unblocked which is causing the signal to never be delivered to the handlers in test-signal-handler.js and it loops until it times out. Not sure if this is the best way of fixing this but hopefully more eyes on this will help. PR-URL: nodejs#12266 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl> Reviewed-By: Anna Henningsen <anna@addaleax.net>

danbev · 2017-04-11T06:43:59Z

@italoacasas I've added #12332 for this and two other commits needed to land this on the v7.x-staging branch.

gibfahn · 2017-06-18T00:05:52Z

Could you retarget #12332 at v6.x-staging (or close it and open a new one)? If yes please follow the guide and raise a backport PR, if no let me know or add the dont-land-on label.

nodejs-github-bot added the c++ Issues and PRs that require attention from people who are familiar with C++. label Apr 7, 2017

bnoordhuis requested changes Apr 7, 2017

View reviewed changes

bnoordhuis approved these changes Apr 7, 2017

View reviewed changes

mscdex added inspector Issues and PRs related to the V8 inspector protocol process Issues and PRs related to the process subsystem. labels Apr 7, 2017

addaleax approved these changes Apr 9, 2017

View reviewed changes

danbev force-pushed the test-signal-handler-failure branch from a2b5d5e to d9d1164 Compare April 9, 2017 17:04

danbev closed this Apr 10, 2017

danbev deleted the test-signal-handler-failure branch April 10, 2017 05:23

italoacasas added the backport-to-v7.x label Apr 10, 2017

jasnell mentioned this pull request May 11, 2017

8.0.0 Release Proposal #12220

Closed

sam-github removed the backport-requested-v7.x label Jun 5, 2017

gibfahn mentioned this pull request Jun 15, 2017

Auditing for 6.11.1 nodejs/Release#230

Closed

3 tasks

gibfahn added the backport-requested-v6.x label Jun 18, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

src: only block SIGUSR1 when HAVE_INSPECTOR #12266

src: only block SIGUSR1 when HAVE_INSPECTOR #12266

danbev commented Apr 7, 2017

danbev commented Apr 7, 2017

bnoordhuis left a comment

bnoordhuis Apr 7, 2017

danbev commented Apr 7, 2017 •

edited

Loading

bnoordhuis commented Apr 7, 2017

danbev commented Apr 7, 2017

danbev commented Apr 7, 2017

bnoordhuis Apr 7, 2017

danbev Apr 7, 2017

danbev commented Apr 9, 2017

danbev commented Apr 10, 2017

italoacasas commented Apr 10, 2017

danbev commented Apr 11, 2017

gibfahn commented Jun 18, 2017

src: only block SIGUSR1 when HAVE_INSPECTOR #12266

src: only block SIGUSR1 when HAVE_INSPECTOR #12266

Conversation

danbev commented Apr 7, 2017

Checklist

Affected core subsystem(s)

danbev commented Apr 7, 2017

bnoordhuis left a comment

Choose a reason for hiding this comment

bnoordhuis Apr 7, 2017

Choose a reason for hiding this comment

danbev commented Apr 7, 2017 • edited Loading

bnoordhuis commented Apr 7, 2017

danbev commented Apr 7, 2017

danbev commented Apr 7, 2017

bnoordhuis Apr 7, 2017

Choose a reason for hiding this comment

danbev Apr 7, 2017

Choose a reason for hiding this comment

danbev commented Apr 9, 2017

danbev commented Apr 10, 2017

italoacasas commented Apr 10, 2017

danbev commented Apr 11, 2017

gibfahn commented Jun 18, 2017

danbev commented Apr 7, 2017 •

edited

Loading