[John Thacker <johnthacker@xxxxxxxxx>]

Another option, which should work with the existing API, is to have a wmem_tree with keys the upper 32 bits whose data nodes are also trees, with keys the lower 32 bits. (You store it by doing an exact wmem_tree_lookup32() match, adding a new tree if not found.) Then you do a normal wmem_tree_lookup32_le() on the upper 32 bits, get the result a tree, and depending on whether the key that matched result was equal or less than equal to the upper 32 bits, call wmem_tree_lookup32_le() on the returned tree with key either the lower 32 bits or on G_MAXUINT32 (the latter will get you the highest result in the tree, for the less than equal match case.)

John

On Tue, Oct 19, 2021 at 12:21 PM qiangxiong.huang via Wireshark-dev <wireshark-dev@xxxxxxxxxxxxx> wrote:

John Thacker,

Thank you for your information. I may try to add xxx_insert64/lookup64().

------------------ Original ------------------

From: "Developer support list for Wireshark" <johnthacker@xxxxxxxxx>;

Date: Tue, Oct 19, 2021 10:42 PM

To: "Developer support list for Wireshark"<wireshark-dev@xxxxxxxxxxxxx>;

Subject: Re: [Wireshark-dev] Is this the bug of wmem_tree_lookup32_array_le()?

On Sun, Oct 17, 2021 at 5:41 AM qiangxiong.huang via Wireshark-dev <wireshark-dev@xxxxxxxxxxxxx> wrote:

Who knows that the current behavior of the wmem_tree_lookup32_array_le() is designed in this way or it is just a bug?

If I want to fix the issue #17633, should I modify this wmem_tree_lookup32_array_le() directly or write a new function like wmem_tree_lookup32_array_le_key_as_big_num() (or other name you recommend)?

While I fully understand why you would want a lookup that would be a lexicographical sort with the order of the 32 bit integers in the key being rank (in order to handle a 64 bit integer as two 32 bit integers), the current dissectors that use the API depend on the current behavior and would break, e.g. https://gitlab.com/wireshark/wireshark/-/blob/master/epan/dissectors/packet-dns.c#L4184-L4185 and https://gitlab.com/wireshark/wireshark/-/blob/master/epan/dissectors/packet-usb.c#L3560-L3574

Generally these dissectors have several independent keys, and what they want to match is an exact match on some keys (transaction ID, for example) and a less than match on others (e.g., frame number.) They do the less than or equal match and then check for strict equality on the keys that need that. Handling the lookup like a lexicographical sort would give incorrect results for those dissectors; the current behavior is as intended.

So if you're going to do it, you need a new API function. But perhaps what you really want is to implement something like wmem_tree_lookup64 that supports 64 bit integers directly rather than expecting the callers to separate into the upper and lower 32 bits. It is possible to do lookups of other types (there's a string lookup using a string comparison function, which might be worth looking at since the method for 32 bit integers calls the GUINT_TO_POINTER macros and is guaranteed to work for 64 bit integers.)

John Thacker

___________________________________________________________________________
Sent via:    Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>
Archives:    https://www.wireshark.org/lists/wireshark-dev
Unsubscribe: https://www.wireshark.org/mailman/options/wireshark-dev
             mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe