Interesting bug

[tleslie]

We have been having mostly good experience with phalanger,
but the issue of (on linux and apache), having to have the phalanger
web site path start with .../phalanger/ is still an issue (any progress)?
but today, we experienced a very interesting bug ...
on windows - no problem, move the dll's and php files over to linux
(compiled dll's from windows vs2010 env), and the site works,
but on web pages (*.php) that we push data to jqGrid, etc, where we did
it with local data push to javascript array via phalanger to php,
when phalanger renders the page, if the page size exceeds a certain amount,
it corrupts the page, often repeating part of the page again, etc,
some kind of buffer run around, and numeric size overrun?
if we reduce the data we are "griding" as it get smaller, the issue
stops, add more data, it corrupts page again. So the phalanger code on linux/mono
has some issue when a page gets to big. I just installed the latest phalanger
as well to be sure it wasn't a bug that got fixed. To reproduce? just
have a php loop that spits out 100-1000's line of javascript assignment , and
kapow!!! (but again, odd, works fine with these 1000's lines in windows).

also, the latest phalanger (& tools) it fixed the issue with errors on
properties missmatched between object in the view (of phalanger),
but the intellisense still shows properties for objects (intellisensing)
that don't belong to it. So i see all the properties for all the objects
i use in that page coming up for any object.


-tl

[Jakub Misek]

Do you have gzip output compression turned on? (ob_gzhandler in Phalanger does not support chunking yet, if you configured smaller buffer size, it may corrupt the output encoding?)

[tleslie]

i am not sure now, we changed the
page to use xml feed to jqGrid,
and now page is not large, but we see the data still
corrupting even thou we push it out not as a php page
(using phalanger as view engine), so it appears to not be
phalanger, but its weird as if its mono ?? its serious bug
you would think everyone would run into, so still looking at it.

-tl

[Jakub Misek]

If it is ok on Windows (.NET/IIS) and wrong on Mono (mod_mono/Apache), then yes. I would guess some configuration (compression, encoding, output encoding, page encoding, file encoding, caching, ...)