Addition of the new flag "file-read-mode" into the analysis server to fix the offset bug with Intel…

pkg/analysis_server/doc/api.html

Index: pkg/analysis_server/doc/api.html
diff --git a/pkg/analysis_server/doc/api.html b/pkg/analysis_server/doc/api.html
index c3214a9c54a7ea1416aaa17f152e8f2be384ff55..c48e11182acf7c54d4fd6a9345ba4480949dc1a8 100644
--- a/pkg/analysis_server/doc/api.html
+++ b/pkg/analysis_server/doc/api.html
@@ -173,6 +173,15 @@ dt.typeDefinition {
       errors produced for all files in the actual analysis roots.
       </dd>
     </dl>
+    <dl>
+      <dt>--file-read-mode</dt>
+      <dd>
+      An enumeration of the ways files can be read from disk. Some clients
+      normalize end of line characters which would make the file offset and
+      range information incorrect. The default option is <tt>as-is</tt>, but
+      can also be set to <tt>normalize-eol</tt>.

[Paul Berry, 2015/01/20 19:10:53]

This documentation should precisely define how normalize-eol behaves, since (as
we observed by looking at emacs) different editors normalize eol's in different
ways.  In particular, it should be clear that:

- '\r\n' is converted to '\n'.
- '\r' by itself is converted to '\n'.
- this happens regardless of the OS editor is running on.

Also, given what we learned about emacs, you might want to consider renaming
"normalize-eol" to something like "normalize-eol-always", since in the future
we'll need to add a mode for emacs that normalizes only normalizes some eol's
under certain specific conditions.

[jwren, 2015/01/20 21:38:14]

Done.

+      </dd>
+    </dl>
     </blockquote>
     <h2 class="domain"><a name="domain_server">Domain: server</a></h2>
       <p>