Re: 4.0.3RC1 reports on *BSD, Alpha and ia64: ok overall

At 01:20 PM 10/3/00 +0200, Olivier Cahagne wrote:

Andi Gutmans wrote: I think SCRIPT_FILENAME should be SCRIPT_NAME. Andi At 12:37 PM 10/3/00 +0300, Andi Gutmans wrote:

Can you try and edit cgi_main.c line 315 and change: SG(request_info).request_uri = getenv("PATH_INFO"); to: SG(request_info).request_uri = getenv("PATH_INFO"); if (!SG(request_info).request_uri || !SG(request_info).request_uri[0]) { SG(request_info).request_uri = getenv("SCRIPT_FILENAME"); } I don't know if this is supposed to fix it but it might.

Fixed, on my RH6.2 setup with --enable-discard-path and on my NetBSD setup too, also with --enable-discard-path.

I commited the patch to the CVS (without the second part of the if() statement which really shouldn't be needed).

It's indeed SCRIPT_NAME, it gives me: <img src="/cgi-bin/info.cgi?=PHPE9568F34-D428-11d2-A769-00AA001ACF42" SCRIPT_FILENAME would give me: <img src="/opt/apache/cgi-bin/info.cgi?=PHPE9568F34-D428-11d2-A769-00AA001ACF42" And Netscape 4.x on Unix display images. ok, my PHP setup here is nearly perfect ! I just need cURL to compile with PHP and, then... I'll start thinking of Sablotron. This needs to be tested on Win32 and in normal cgi mode in Unix. I really don't know if it should be in 4.0.3, images not displaying is no so important and it's the only way this bug showed up, I don't know if the fix affects cgi's behaviour in another way. Shouldn't the fix be part of the #if DISCARD_PATH as the bug only showed when using --enable-discard-path ?

I don't think this patch can do any harm. It uses SCRIPT_NAME as the URI only if PATH_INFO is not defined. Andi --- Andi Gutmans <andi@zend.com> http://www.zend.com/