Cross-site scripting (XSS) vulnerability in results.php in BrowserCRM allows remote attackers to inject arbitrary web script or HTML via certain manipulations of the query parameter, as demonstrated using an IMG SRC tag.

Opera 8.02 and earlier allows remote attackers to cause a denial of service (client crash) via (1) a crafted HTML file with a "content: url(0);" style attribute, a "bodyA" tag, a long string, and a "u" tag with a long attribute, as demonstrated by opera.html; and (2) a BGSOUND element with a "margin:-99;" STYLE attribute.

Opera before 8.51, when running on Windows with Input Method Editor (IME) installed, allows remote attackers to cause a denial of service (persistent application crash) by bookmarking a site with a long title.

Opera 8.50 allows remote attackers to cause a denial of service (crash) via a Java applet with a large string argument to the removeMember JNI method for the com.opera.JSObject class.

Opera before 8.51 on Linux and Unix systems allows remote attackers to execute arbitrary code via shell metacharacters (backticks) in a URL that another product provides in a command line argument when launching Opera.

Opera Web Browser 8.50 and 8.0 through 8.0.2 allows remote attackers to spoof the URL in the status bar via the title in an image in a link to a trusted site within a form to the malicious site.

Multiple unspecified vulnerabilities in Opera 8.50 on Linux and Windows have unknown impact and attack vectors, related to (1) " handling of must-revalidate cache directive for HTTPS pages" or (2) a "display issue with cookie comment encoding."

Unspecified "drag-and-drop vulnerability" in Opera Web Browser before 8.50 on Windows allows "unintentional file uploads."

The mail client in Opera before 8.50 opens attached files from the user's cache directory without warning the user, which might allow remote attackers to inject arbitrary web script and spoof attachment filenames.

Opera before 8.50 allows remote attackers to spoof the content type of files via a filename with a trailing "." (dot), which might allow remote attackers to trick users into processing dangerous content.