couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benoit Chesneau" <>
Subject Re: Call for testing: couchjs branch
Date Tue, 04 Mar 2008 12:35:48 GMT
On Mon, Mar 3, 2008 at 11:20 AM, Christopher Lenz <> wrote:
> Hey all,
>  we've been working on a branch that would remove the SpiderMonkey
>  sources from the CouchDB repository, so that we could instead treat
>  SpiderMonkey as a normal external library dependency.
>  See <
>   > for some background.
>  So what we've done on the "couchjs" branch is to add a simple C
>  wrapper around the SpiderMonkey lib (src/CouchDB/couch_js.c) that
>  implements only those parts of the JS shell that we actually need for
>  the view server, with the minor modifications we need. And of course
>  the build system was adjusted so that SpiderMonkey is detected as an
>  external dependency.
>  The branch has now come so far along that it "worksforme", so I'd like
>  to ask anyone who's interested to check out the branch, do the build
>  and run some tests, and report back problems (or, hopefully at least
>  in some cases, the absence of problems ;).

tested on openbsd 4.3-current on amd64. The bootstrap don't work,
mostly because spidermonkey is compiled with threads support and it
don't find -ljs. Find patchs enclosed.

After having installed I tested it and have following errors :
- conflicts : Run with debuggerAssertion failed: e.error == "conflict"
- replication : # Exception raised:
[{couch_rep,do_http_request,3},\n {couch_rep,get_missing_revs,2},\n
{couch_rep,maybe_save_docs,4},\n {couch_rep,'-pull_rep/4-fun-0-',5},\n
{couch_rep,enum_docs0,3},\n {couch_rep,enum_docs_since,4},\n
{couch_rep,pull_rep,4},\n {couch_rep,replicate,3}]}"}

Anotherthing, and it is true for 0.7.2 too, couchdb is very slow on
openbsd, i attached a screenshot of time results in test suite. Same
machine on linux and i have time 90% faster... (first test : 400 msecs
on linux against 6400 on openbsd). I don't understand why. Any idee
how I could test it further ?


- benoƮt

  • Unnamed multipart/mixed (inline, None, 0 bytes)
View raw message