Handle Z_BUF_ERROR as a non-fatal error while incrementally decompressing

Handle Z_BUF_ERROR as a non-fatal error while incrementally decompressing
SYN_STREAM/SYN_REPLY headers.

This also fixes bug when server sends SYN_REPLY in separate chunks
(i.e. header + zlib chunks ) chrome fails to recognize it.

BUG=108596
R=jar
TEST=ASAN tests and network unit tests
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=117563

[ramant (doing other things), 2012-01-12 01:13:36]

Hi Jim,
  Added the changes from spdy_framer server code to handle zlib's Z_BUF_ERROR. 

  With this fix, the test (to simulate ASAN test failure) when rand() returned 0
always worked. Will land this change in the night and will revert if ASAN tests
fail.

  We keep spdy_framer.cc and spdy_framer_test.cc in sync with the server code
(took their change as is without making any additional changes).

  Would appreciate your comments.

thanks much,
raman

[jar (doing other things), 2012-01-12 23:11:22]

lgtm

[wtc, 2012-01-12 23:49:38]

LGTM.

http://codereview.chromium.org/9185036/diff/1/net/spdy/spdy_framer.cc
File net/spdy/spdy_framer.cc (right):

http://codereview.chromium.org/9185036/diff/1/net/spdy/spdy_framer.cc#newcode...
net/spdy/spdy_framer.cc:1529: if (rv != Z_OK && rv != Z_BUF_ERROR) {

Z_BUF_ERROR means "no progress is possible".  With that
knowledge, we may be able to do something simpler.

[commit-bot: I haz the power, 2012-01-13 00:13:36]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/rtenneti@chromium.org/9185036/1

[ramant (doing other things), 2012-01-13 00:14:58]

http://codereview.chromium.org/9185036/diff/1/net/spdy/spdy_framer.cc
File net/spdy/spdy_framer.cc (right):

http://codereview.chromium.org/9185036/diff/1/net/spdy/spdy_framer.cc#newcode...
net/spdy/spdy_framer.cc:1529: if (rv != Z_OK && rv != Z_BUF_ERROR) {
On 2012/01/12 23:49:38, wtc wrote:
> 
> Z_BUF_ERROR means "no progress is possible".  With that
> knowledge, we may be able to do something simpler.

Wan-Teh and I had stepped through the code and we are satisfied with the fix. It
exits the function because decomp->avail_in is zero.

Will discuss with server group if we could rearrange this code so that it is
more readable.

Done.