Hello Zack,
1) I would like to ask your permission to submit the whole code for code submission. First because my task in the summer wasn't only to write the parts that I wrote but also to read the code and debug it as well. Secondly because the git log shows who wrote what. Thirdly, It's easier.
2) I had more discussion on libcurl-dev with the libcurl main developer. He thinks despite the fact my approach is working right now, it isn't stable and can break down in a different environment. To solve this, I'm going to let curl also read from the socket instead of libevent, this isn't very hard (I'll disable both EV_READ and EV_WRITE as soon as I'm handing the socket to curl). I confirmed with Nick that (unlike libcurl) libevent support this.
3) Thank you for telling me about fts. I'm going to replace boost code with fts soon.
4)
I still don't see the problem. If you do exactly that, and don't use he DECLARE macros for http_apache_steg, what breaks?
If I don't use the DECLARE then I can't specify http_apache as the steg protocol in the command line .
Best, vmon
On Mon, Sep 10, 2012 at 9:10 AM, Ian Goldberg iang@cs.uwaterloo.ca wrote:
On Mon, Sep 10, 2012 at 02:33:30AM -0600, vmon wrote:
- Thank you for telling me about fts. I'm going to replace boost code with fts soon.
What is fts? This sounds potentially useful.
It's a C API for directory tree walking. Similar to the older ftw() but iterator-style and more robust. Unfortunately doesn't seem to have made it into POSIX yet but is available on both Linux and OSX.
http://www.kernel.org/doc/man-pages/online/pages/man3/fts.3.html
On 2012-09-10 16:58 , Zack Weinberg wrote:
On Mon, Sep 10, 2012 at 9:10 AM, Ian Goldberg iang@cs.uwaterloo.ca wrote:
On Mon, Sep 10, 2012 at 02:33:30AM -0600, vmon wrote:
- Thank you for telling me about fts. I'm going to replace boost code with fts soon.
What is fts? This sounds potentially useful.
It's a C API for directory tree walking. Similar to the older ftw() but iterator-style and more robust. Unfortunately doesn't seem to have made it into POSIX yet but is available on both Linux and OSX.
And even OpenBSD has it ;)
Greets, Jeroen
On Mon, Sep 10, 2012 at 4:33 AM, vmon vmonmoonshine@gmail.com wrote:
Hello Zack,
- I would like to ask your permission to submit the whole code for
code submission. First because my task in the summer wasn't only to write the parts that I wrote but also to read the code and debug it as well. Secondly because the git log shows who wrote what. Thirdly, It's easier.
Yes, that seems reasonable. In case Google decides they don't like that, you could also send them the diff between master and your branch.
- I had more discussion on libcurl-dev with the libcurl main developer. He
thinks despite the fact my approach is working right now, it isn't stable and can break down in a different environment. To solve this, I'm going to let curl also read from the socket instead of libevent, this isn't very hard (I'll disable both EV_READ and EV_WRITE as soon as I'm handing the socket to curl). I confirmed with Nick that (unlike libcurl) libevent support this.
I don't know enough about libcurl to really grok the situation here, but what you're doing sounds plausible.
I still don't see the problem. If you do exactly that, and don't use he DECLARE macros for http_apache_steg, what breaks?
If I don't use the DECLARE then I can't specify http_apache as the steg protocol in the command line .
Ah, so what you need to do is define a *new* DECLARE macro that doesn't define all the boilerplate you don't want - let's call it STEG_DECLARE_SUBMODULE - and modify genmodtable.sh to pick it up.
zw
-
Ian Goldberg -
Jeroen Massar -
vmon -
Zack Weinberg