{"id":9323,"date":"2011-10-21T07:00:00","date_gmt":"2011-10-21T07:00:00","guid":{"rendered":"https:\/\/blogs.msdn.microsoft.com\/oldnewthing\/2011\/10\/21\/the-psn_setactive-notification-is-sent-each-time-your-wizard-page-is-activated\/"},"modified":"2011-10-21T07:00:00","modified_gmt":"2011-10-21T07:00:00","slug":"the-psn_setactive-notification-is-sent-each-time-your-wizard-page-is-activated","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/oldnewthing\/20111021-00\/?p=9323","title":{"rendered":"The PSN_SETACTIVE notification is sent each time your wizard page is activated"},"content":{"rendered":"

\nA customer had received a number of crashes via\n\nWindows Error Reporting<\/a>\nand believed that they had found a bug in the tree view common control.\n<\/p>\n

\n

\nIn our UI, we have a tree view with checkboxes.\nThe tree view displays a fixed item at the top,\nfollowed by a variable number of dynamic items.\nWhen the user clicks Next<\/i>,\nwe look at the tree view to determine what the user selected.\nThe code goes like this (pseudo):\n<\/p>\n

\nhtiRoot = GetTreeRoot();\n\/\/ First process the fixed item\nhtiFixed = GetChild(htiRoot);\nif (IsTreeViewItemChecked(htiFixed)) {\n    .. add the fixed item ...\n}\n\/\/ Now process the dynamic items\nhti = GetNextSibling(htiFixed);\nwhile (hti != NULL) {\n  if (IsTreeViewItemChecked(hti)) {\n    ... add the dynamic item ...\n  }\n  hti = GetNextSibling(hti);\n}\n<\/pre>\n
\nIn the crashes we receive, other variables in the\nprogram indicate that there should be only one dynamic item,\nbut our loop iterates multiple times.\nFurthermore, the first time through the loop,\nthe hItem<\/code> is not the handle to the first\ndynamic item but is in fact the handle to the fixed item.\nThis naturally results in a crash when we try to treat the fixed item\nas if it were a dynamic item.\n<\/p>\n
\nAnother thing we noticed is that at the time of the crash,\nall three variables htiRoot<\/code>\nhtiFixed<\/code>,\nand\nhti<\/code> have the same value.\n<\/p>\n
\nOur attempts to reproduce the problem in-house have been\nunsuccessful.\nFrom our analysis, we believe that the tree view APIs used to\nobtain handles to children and sibling nodes are misbehaving.\n<\/p>\n<\/blockquote>\n
\nThe customer included the crash bucket number,\nso we were able to connect to the same crash dumps that the customer\nwas looking at.\n<\/p>\n
\nThe first thing to dismiss was the remark that all three of\nthe local variables had the same value.\nThis is to be expected since they have non-overlapping lifetimes,\nand the compiler decided to alias them all to each other to save\nmemory.\n<\/p>\n
\n...\n        lea     eax,[ebp+8]         ; htiRoot\n        push    eax\n        push    1                   ; some flag\n        push    ebx                 ; some parameter\n        call    00965fb9            ; GetTreeRoot\n        mov     [ebp-2Ch],eax\n        test    eax, eax\n        jl      00971a49            ; failed - exit\n        mov     edi, [_imp__SendMessageW]\n        push    4                   ; TVGN_CHILD\n        push    110Ah               ; TVM_GETNEXTITEM\n        push    dword ptr [ebx+10h] ; window handle\n        call    edi                 ; SendMessage\n        mov     [ebp+8],eax         ; htiFixed\n    ... eliding if (IsTreeViewItemChecked(...)) ...\n        jmp     00971a1c            ; enter loop\n00971931:\n    ... eliding if (IsTreeViewItemChecked(...)) ...\n00971a1c:\n        push    dword ptr [ebp+8]   ; hti\n        push    1                   ; TVGN_NEXT\n        push    110Ah               ; TVM_GETNEXTITEM\n        push    dword ptr [ebx+10h] ; window handle\n        call    edi                 ; SendMessage\n        mov     [ebp+8],eax         ; update hti\n        test    eax, eax            ; hti == NULL?\n        jne     00971931            ; N: continue loop\n<\/pre>\n
\nI’ve removed code not directly relevant to the discussion.\nThe point to see here is that the compiler combined all three\nvariables into one physical memory location at\n[ebp+8]<\/code>\nsince there is no\npoint in the program where more than one of the values is needed\nat a time.\nIn other words, the compiler rewrote your code like this:\n<\/p>\n
\nhti = GetTreeRoot();\nhti = GetChild(hti);\nif (IsTreeViewItemChecked(hti)) {\n    .. add the fixed item ...\n}\nwhile ((hti = GetNextSibling(hti)) != NULL) {\n  if (IsTreeViewItemChecked(hti)) {\n    ... add the dynamic item ...\n  }\n}\n<\/pre>\n
\nNot only did the compiler merge all your hti<\/code>\nvariables into one, it realized that once it did that,\nthe two calls to Get­Next­Sibling<\/code>\ncould be folded together as well.\n<\/p>\n
\nOkay, one mystery solved.\nWhat about the others?\n<\/p>\n
\nFrom studying the crash dump, the shell team determined that\nthe reason the first dynamic item appears to be the fixed item\nis that the tree view actually has two<\/i> fixed items:\n<\/p>\n
\n003d06d8 Root\n+ 003d0a38 \"Configuration settings\"\n+ 003d0888 \"Configuration settings\"\n+ 003d07b0 \"Saved game from May 27, 2009 at 2:42 PM (playing as Thor)\"\n+ 003d0600 \"Saved game from May 27, 2009 at 2:42 PM (playing as Thor)\"\n<\/pre>\n
\n“Configuration settings” is the fixed item, and the saved\ngames are the dynamic items.\n(This isn’t the actual scenario from the customer, but it\ngets the point across.)\nThe customer was wrong to use the definite article when referring\nto the<\/i> handle to\nthe<\/i> fixed item, since there are two fixed items here.\nIn a sense, the customer’s understanding that there is only\none fixed item clouded their ability to debug the problem:\nWhen they saw another fixed item, they assumed not that they\nreceived another item that was fixed,\nbut rather that they were getting the same fixed item twice.\n<\/p>\n
\nSeeing that the tree view was being populated twice\ndirected the next step of the investigation: Why?\n<\/p>\n
\nThe code that populates the tree view is called from\nthe wizard page’s PSN_SET­ACTIVE<\/code> notification,\nand that one piece of information was the last piece of the\npuzzle.\n<\/p>\n
\nThe PSN_SET­ACTIVE<\/code> notification is sent each time\nthe wizard or property sheet page is selected as the current page.\nIf the page is activated twice, then you will get two\nPSN_SET­ACTIVE<\/code> notifications.\nThe solution was to populate the tree view only the first\ntime the page was activated.\n<\/p>\n
\nExercise<\/b>: What was missing from the customer’s testing\nthat prevented them from reproducing the problem in their labs?<\/p>\n","protected":false},"excerpt":{"rendered":"
A customer had received a number of crashes via Windows Error Reporting and believed that they had found a bug in the tree view common control. In our UI, we have a tree view with checkboxes. The tree view displays a fixed item at the top, followed by a variable number of dynamic items. When […]<\/p>\n","protected":false},"author":1069,"featured_media":111744,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[1],"tags":[25],"class_list":["post-9323","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-oldnewthing","tag-code"],"acf":[],"blog_post_summary":"
A customer had received a number of crashes via Windows Error Reporting and believed that they had found a bug in the tree view common control. In our UI, we have a tree view with checkboxes. The tree view displays a fixed item at the top, followed by a variable number of dynamic items. When […]<\/p>\n","_links":{"self":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts\/9323","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/users\/1069"}],"replies":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/comments?post=9323"}],"version-history":[{"count":0,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/posts\/9323\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/media\/111744"}],"wp:attachment":[{"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/media?parent=9323"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/categories?post=9323"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/oldnewthing\/wp-json\/wp\/v2\/tags?post=9323"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}