HTTP cache: Add tests to improve coverage of AddTruncatedFlag.

net/http/http_cache_transaction.cc

Index: net/http/http_cache_transaction.cc
diff --git a/net/http/http_cache_transaction.cc b/net/http/http_cache_transaction.cc
index 450cc450eebc45ce14a61c59e0b4b301202c4f1b..db3f4968911b47ec29167190cf429a69007e00d7 100644
--- a/net/http/http_cache_transaction.cc
+++ b/net/http/http_cache_transaction.cc
@@ -1875,7 +1875,7 @@ int HttpCache::Transaction::DoCacheWriteDataComplete(int result) {
     // We want to ignore errors writing to disk and just keep reading from
     // the network.
     result = write_len_;
-  } else if (!done_reading_ && entry_) {
+  } else if (!done_reading_ && entry_ && (!partial_ || truncated_)) {

[asanka, 2015/09/17 14:43:34]

I'll note that this change doesn't affect the newly added tests. Perhaps with
this change, we can make done_reading_ be a stronger guarantee so that once the
SM moves past this point, there's no longer a need to check for partial_ or
sparse_ to decide whether to mark the entry as truncated.

[rvargas (doing something else), 2015/09/17 20:18:37]

These tests are actually a regression test for
https://codereview.chromium.org/1253173002/ (I believe you based your CL between
the regression and the fix).

One of the tests I'm adding doesn't fail without this line... the other fails
all the time.

There's certainly room for improvement in the meaning of the flags, and when
they are valid :). What I'm saying with this line is that (so far), byte range
request imply specific logic to figure out when we are done, not what we do here
for done_reading_ (which is quite permissive, due to buggy servers).

     int current_size = entry_->disk_entry->GetDataSize(kResponseContentIndex);
     int64 body_size = response_.headers->GetContentLength();
     if (body_size >= 0 && body_size <= current_size)