-
Notifications
You must be signed in to change notification settings - Fork 29.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[wip] merge v0.10 #267
[wip] merge v0.10 #267
Conversation
@@ -216,6 +219,11 @@ email.md: ChangeLog tools/email-footer.md | |||
blog.html: email.md | |||
cat $< | ./$(NODE_EXE) tools/doc/node_modules/.bin/marked > $@ | |||
|
|||
doc-branch-upload: NODE_DOC_VERSION = v$(shell $(PYTHON) tools/getnodeversion.py | cut -f1,2 -d.) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@bnoordhuis removed upload targets in 73b9323, which I'm +1 on, I'll either PR to add some new ones back or we'll keep them separate. We should keep that consistent and remove this target.
8848675
to
44ddf76
Compare
Regarding the SSL/TLS commits, only the first two and last three seem worth taking. The others are all about making SSLv[23] support configurable at run-time and I don't think that's a path we want to take. Personally, I would suggest ripping out SSLv2 support entirely and putting SSLv3 behind a configure flag that prints a big warning. /cc @indutny Agree/disagree? |
+1, what was the argument for keeping SSLv2? IE6? |
@rvagg I don't know, it's probably pure inertia. SSLv2 was considered broken almost immediately after its introduction in 1995, that's why SSLv3 was created shortly after. There really seems no reason to keep the support around. |
@piscisaureus The uncaught exception changes seem pointless, they work around an issue when --abort_on_uncaught_exception is passed on the command line, effectively nullifying the flag. When you tell a GP 'it hurts when I do this', they'll tell you to stop doing that. Tell a programmer, they'll add a hack. |
@piscisaureus The timer commits can be omitted when some of the suggestions from #268 get implemented (because that practically constitutes a full rewrite, nullifying the issues those commits address.) |
@bnoordhuis Thanks for your replies. I agree with most of them except
It's not that I think using SSLv3 is a good idea, but I'm generally opposed to the idea of putting features behind a configure-time flag. It assumes that people build iojs themselves, which generally speaking only a very small fraction of the user base does; most people will get it from a package manager or installer. So we should choose: we either remove support for SSLv3 entirely, or we make a run-time decision possible. Let me add that although you've indeed removed SSLv2 support a year ago, this change never saw the light of day in the form of a stable release. |
I'm in the 'secure by default' camp, +1 for dropping SSLv3 support. @iojs/tc, please chime in. |
+1 for dropping SSLv3 from me |
@bnoordhuis re: the uncaught exception handler: there was a specific issue that this was handling -- specifically, I think it was that using domains at the same time as |
Re: disabling SSLv3 entirely: I think this is a good opportunity to turn it off by default. If a lot of folks complain, we can go the runtime configuration route in a follow-up minor release. I'm -1 on a configure flag, for the same reason as @piscisaureus. |
@chrisdickinson Re: --abort-on-uncaught-exception, I think a much easier - and better - solution is to tell users to not use that flag. I can put together a PR that makes iojs print a warning on start-up or
|
I think it won't hurt to land these TLS patches: |
@indutny @piscisaureus I mentioned up-thread that the first two and last three may be worth cherry-picking. nodejs/node-v0.x-archive@b9283cf is not because we won't be merging the SSL2_ENABLE/SSL3_ENABLE patches, right? |
Ok, good! |
Add a test that goes through the whole matrix of: - command line options (--enable-ssl*) - secureOptions - secureProtocols and makes sure that compatible test setups actually work as expected. The test works by spawning two processes for each test case: one client and one server. The test passes if a SSL/TLS connection from the client to the server is successful and the test case was supposed to pass, or if the connection couldn't be established and the test case was supposed to fail. The test is currently located in the directory 'test/external' because it has external dependencies. Cherry-picked-from: nodejs/node-v0.x-archive@8d045a3
Cherry-picked-from: nodejs/node-v0.x-archive@641cea0 Conflicts: Makefile
44ddf76
to
64086a9
Compare
@piscisaureus tbh I don't really see the utility, it's providing a version |
actually this is an issue for the new docs team to take up! I delegate the decision to them, whoever they be / @chrisdickinson |
@bnoordhuis Re: abort-on-uncaught-exception, I think that approach might preclude anyone using hapi from making use of post-mortem debugging. Though, this may not be a problem if there are plans to rip domains out of hapi -- cc'ing @geek to see if that's the plan or not. @rvagg I'm not entirely sold on including that commit, either. I'm currently leaning towards giving the doctool responsibility for the entirety of the |
Is this still in progress? |
Many patches have been ported over from the node/v0.10 branch onto io.js. Others are in this pull request. The following still need to be looked at.
Once all these commits have either been ported or rejected, io.js is up-to-date up to nodejs/node-v0.x-archive@fe20196 and we can construct a new merge base.
Child process fixes: (@sam-github @cjihrig)
TLS related (@indutny @bnoordhuis)
Uncaught exceptions swallowed
Timers (see also #268)
Probably not needed
Misc (@piscisaureus)