Hi Yes, it could be considered a bug. But it looks strange to set it on a hidden field. Anyway, the argumentation is solid, so please create an issue. It could be great if you can say in which browsers this behavior happens.
regards, Leonardo Uribe 2015-01-13 14:31 GMT-05:00 Werner Punz <[email protected]>: > Am 09.01.15 um 11:58 schrieb Dan Østerberg: > >> Hi, >> >> We have a JSF application where we mainly use ViewScoped beans, and add >> no-cache headers for all JSF pages. Navigating back to a previous page >> correctly recreates the beans, and renders new HTML, with a new ViewState >> ID. But... some browsers "autocomplete" some of the hidden ViewState inputs, >> overriding the new value with an old ViewState value. We have verified this >> in the browser dev-tools by looking at the response, which is correct, and >> the resulting HTML which is not. >> >> In short, this is a known autocomplete issue, which Mojarra has fixed >> since 1.2, by adding 'autocomplete="off"' to the hidden ViewState input. >> Plus a context parameter "com.sun.faces.autoCompleteOffOnViewState" for >> opting not to use it, since it results in invalid XHTML. Adding >> 'pa:autocomplete="off"' explicitly to the whole form also fixes this issue. >> However, at least the MyFaces version that we use (2.2.4) doesn't add this >> attribute, and doesn't seem to have any corresponding configuration either. >> We also failed to google up alternatives/explanations for this issue >> explicitly in MyFaces. Naturally, we would like to avoid javascript hacks >> and custom components and renderers. >> >> So the question is - are we missing something? Or should MyFaces be >> patched to simply render 'autocomplete="off"' for its hidden >> javax.faces.ViewState inputs? >> >> Thanx, >> Dan Østerberg >> > Hi this looks like a bug to me, please file a bugreport in the MyFaces > Bugtracker. > https://issues.apache.org/jira/browse/myfaces > > > Kind regards > > Werner Punz > >
