| Summary: | mod_deflate generates corrupted result at large data. | ||
|---|---|---|---|
| Product: | Apache httpd-2 | Reporter: | ASADA Kazuhisa <kaz> |
| Component: | mod_deflate | Assignee: | Apache HTTPD Bugs Mailing List <bugs> |
| Status: | CLOSED FIXED | ||
| Severity: | normal | CC: | bruno |
| Priority: | P3 | ||
| Version: | 2.0-HEAD | ||
| Target Milestone: | --- | ||
| Hardware: | Sun | ||
| OS: | Solaris | ||
Woo-hoo! I was tracking this down last week, but I had to leave for WWDC and didn't follow up. Your patch is right and has been committed in revision 1.4 of modules/filters/mod_deflate.c. It will be included in the next release of Apache 2.0! Thanks for using Apache (and keep the patches coming)! I am still seeing these symptoms in the httpd-2.0_20020530221254.tar.gz snapshot. The difference may be that for the files having problems the output is generated by a cgi script and then fed through INCLUDES and then DEFLATE. 50KB files work OK. 450KB files fail to be uncompressed using lynx 2.5-dev, Netscape 4.08 and IE 5.?. I am running this under linux 2.2.16 (Redhat) on a Athlon Tbird. I tested things a bit more and found that if I am just using the DEFLATE filter things work OK. But when I use DEFLATE after INCLUDES then they don't. I also tried just INCLUDES and the only differences were the expected ones. Is there any chance that you can provide a sample CGI script that generates the right include tags to causes mod_include to create a sequence of data that mod_deflate can't handle? Thanks! A CGI-BIN script isn't needed to duplicate the problem. A sample broken link is at: http://wolff.to/area/test.html Source to the actual files is available at: http://wolff.to/area/test.txt http://wolff.to/area/sig.txt http://wolff.to/area/links.txt The .txt files are sym links to the .html files. INCLUDES and DEFLATE filters are run on .html files, but not on .txt files. Justin Erenkrantz committed a patch to mod_deflate earlier today that might fix this problem. Try out the CVS version of mod_deflate.c (rev 1.16) and see if that fixes your problem. Thanks! I tried 1.16 in a CVS snapshot from a few days ago and the problem appears to be fixed. I will be removing the test files shortly. Thanks for fixing this. |
mod_deflate module generates corrupted result with large data (over 200kbytes). deflate_out_filter deflate input buckets into output bucket, but output buffer information reset every input bucket, it means waste partial deflated data at process next input bucket. Output buffer must initialize before BRIGADE_FOREACH loop. --- mod_deflate.c.org Tue May 7 09:35:16 2002 +++ mod_deflate.c Mon May 13 00:32:18 2002 @@ -328,4 +328,8 @@ } + /* initialize deflate output buffer */ + ctx->stream.next_out = ctx->buffer; + ctx->stream.avail_out = c->bufferSize; + APR_BRIGADE_FOREACH(e, bb) { const char *data; @@ -437,6 +441,4 @@ * trust zlib */ ctx->stream.avail_in = len; - ctx->stream.next_out = ctx->buffer; - ctx->stream.avail_out = c->bufferSize; while (ctx->stream.avail_in != 0) {