1 |
Dnia 2015-01-27, o godz. 02:46:37 |
2 |
Andrew Savchenko <bircoph@g.o> napisał(a): |
3 |
|
4 |
> Hi, |
5 |
> |
6 |
> On Sun, 25 Jan 2015 14:59:01 +0100 Michał Górny wrote: |
7 |
> > Dnia 2015-01-21, o godz. 11:05:34 |
8 |
> > Michał Górny <mgorny@g.o> napisał(a): |
9 |
> > |
10 |
> > > Generic proxy solution |
11 |
> > > ---------------------- |
12 |
> > > |
13 |
> > > The simplest solution so far seems to be setting a generic SOCKS proxy |
14 |
> > > inside the build environment, and wrapping distcc so that it will use |
15 |
> > > it for network access. |
16 |
> > > |
17 |
> > > Unless we do some extra magic which don't want to do, this means that |
18 |
> > > other apps can also abuse the proxy to reach outside sandbox. However, |
19 |
> > > network-sandbox is not really a security feature, so I don't think that |
20 |
> > > is important. At least as long as we don't export it globally :). |
21 |
> > > |
22 |
> > > Of course, software is a problem. We'd need at least some SOCKS server |
23 |
> > > for Portage (at least a very simple one), and as far as I'm aware |
24 |
> > > distcc does not support SOCKS directly, so tsocks in addition to that. |
25 |
> > |
26 |
> > So finally went this way instead. |
27 |
> |
28 |
> I still don't understand why. This solution: |
29 |
> 1) is intrusive, it requires patching distcc and upstream as good |
30 |
> as dead (see below); |
31 |
|
32 |
Adding SOCKSv5 support is something of potentially generic benefit |
33 |
unlike adding Gentoo-specific hacks. |
34 |
|
35 |
> 2) will require a _separate_ solution for icecream and thus a |
36 |
> double effort; |
37 |
|
38 |
It would require a separate solution for icecream anyway. If icecream |
39 |
had SOCKSv5 support, the solution would be as simple as exporting extra |
40 |
variable for it. Unlike the other case when Portage needs to become |
41 |
aware of layout of another tool. |
42 |
|
43 |
> 3) adds additional latency for distcc network path, which is |
44 |
> undesirable. |
45 |
|
46 |
Unlikely to be measurable. ~20 extra bytes over UNIX socket. In fact, |
47 |
this provides us with ability to do some tricks to actually improve |
48 |
the latency -- like caching extra connections in the proxy. |
49 |
|
50 |
In fact, the other solution is more complex and therefore more likely |
51 |
to cause delays. |
52 |
|
53 |
> Parent namespace solution looks like the most reasonable for me |
54 |
> based on both arguments above and years of heavy distcc usage |
55 |
> experience. |
56 |
> |
57 |
> > [2]:https://code.google.com/p/distcc/issues/detail?id=149 |
58 |
> |
59 |
> Chances to have this upstream are close to zero. If it is not |
60 |
> dead, it is very close to it. Number of bugs and patches is |
61 |
> accumulating without any response. No releases from 2011. |
62 |
> Probably someone should fork it... |
63 |
|
64 |
I know. Forgive me the wording but it's pretty much crap code. |
65 |
Involving ideas such as enabling non-blocking I/O just to read it |
66 |
in loops to get blocking behavior. |
67 |
|
68 |
I would rewrite it from scratch if it were simpler. But the pump mode |
69 |
and other extra features makes it require too much effort. At least |
70 |
it's not something I could do in my free time. |
71 |
|
72 |
> Distcc has a problem with -march=native right now: it just falls |
73 |
> back to local compilation if encountered it. I sent them a patch |
74 |
> 1.5 years ago [1] and still no reply... It also requires some |
75 |
> patches for successful cross-compilation when plain gcc is invoked |
76 |
> by the client. (I have patches for amd64 <-> x86 only and they may |
77 |
> broke pump mode (never used it anyway), thus I haven't send them |
78 |
> upstream. |
79 |
> |
80 |
> [1] https://groups.google.com/forum/#!topic/distcc-patches/eeP-9pTgz7E |
81 |
> |
82 |
> In short, this patch expands "native" argument using gcc output, |
83 |
> caches result (based on fingerprint of compiler being invoked) |
84 |
> and sends expanded string to distcc servers. It is in my overlay |
85 |
> (bircoph) if someone is interested. Works fine for me all these |
86 |
> time. |
87 |
|
88 |
I'm not convinced it's worth the effort to support it in distcc. I once |
89 |
published a blog post how to use gcc output to replace -march=native |
90 |
with the expanded flags. I find that saner. |
91 |
|
92 |
That said, pump mode is the only significant change in late distcc |
93 |
history. And it's broken anyway, and I doubt anybody's going to fix it |
94 |
since the failures are pretty random and hard to reproduce. |
95 |
|
96 |
-- |
97 |
Best regards, |
98 |
Michał Górny |