Mike Percy has posted comments on this change.

Change subject: Ignore SIGPIPE earlier in startup process
......................................................................


Patch Set 1:

(1 comment)

> What's the plan for the client?

I thought we could deal with the client in a separate patch. The idea I had 
there was to add an option to KuduClientBuilder called ignore_sigpipe(bool) 
that defaults to true. If clients want to manage their own signal handling 
themselves, at a process- or thread- level, then they can set it to false when 
constructing the client. If it's set to true, it calls something like 
EnsureSigPipeIgnored().

http://gerrit.cloudera.org:8080/#/c/6262/1/src/kudu/util/signal.cc
File src/kudu/util/signal.cc:

PS1, Line 38:   static GoogleOnceType once = GOOGLE_ONCE_INIT;
            :   GoogleOnceInit(&once, &IgnoreSigPipe);
> I think you just moved this code, but actually curious why we bother using 
Yeah I did just move this code... I assumed it's to avoid a syscall when we 
think we don't need it. I'm a little ambivalent on it and do think avoiding a 
static GoogleOnce is good for testability, so I'm not opposed to removing this.

Actually, I would like to remove the call to this from Subprocess::Start() but 
I'm a little worried about opening us up to additional test flakiness because 
of that. On the other hand, the call there could be masking other issues.


-- 
To view, visit http://gerrit.cloudera.org:8080/6262
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I040bd38ff31451ed9e25e7cf2127c869cf08a628
Gerrit-PatchSet: 1
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Mike Percy <[email protected]>
Gerrit-Reviewer: David Ribeiro Alves <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Mike Percy <[email protected]>
Gerrit-Reviewer: Todd Lipcon <[email protected]>
Gerrit-HasComments: Yes

Reply via email to