Deprecate ROM filename #1787
Comments
When names that contain ' -> ' are encountered at Rom_connection the trailing element is used as the 'filename' argument. Issue genodelabs#1787
Creates session label strings with the form (label, parent_label). Issue genodelabs#1787
|
I am not sure if the same reasoning applies to the root argument of file-system sessions. The root argument allows the client to use knowledge about the file system hierarchy for limiting its own view of the file system. As far as I can see, in contrast to the label, the root argument is not meant as a criterion for taking session-routing decisions or for selecting a policy at the file-system server. |
|
I've started running into problems with ROM session request buffer overruns from large names being duplicated between the label and filename arguments, so I went ahead and removed the filename argument from ROM session. I wouldn't say that this is a priority issue for me, but I think it does make things a bit simplier. The following replaces the previous two commits. |
New String class for session labels. Convience function for extracting Label from Arg_string. Issue genodelabs#1787
Conveying the ROM filename as the final label element simplifies routing policy and session construction. Fixes genodelabs#1787
Additional edits. Issue genodelabs#1787
Fixed unsigned integer rollover. Issue genodelabs#1787
rename "tail" to "last_element" in utilities Issue genodelabs#1787
|
rebase |
New String class for session labels. Convience function for extracting Label from Arg_string. Issue genodelabs#1787
Conveying the ROM filename as the final label element simplifies routing policy and session construction. Fixes genodelabs#1787
|
You're right, we should consider to integrate the cleanup in the next release. |
Fix the linux dataspace filename retrieval. Fix the bomb test. Issue genodelabs#1787
|
I'd like to merge this cleanup. @ehmry could you please squash and rebase these changes on current master? |
|
Rebased against master. fba1454 is dangerous because it changes the result of ''Session_label(args)'', but I think it is worth it to decouple Session_label from Session_args. I'm adjusting my Nix work to run against master, and I'll need this to clear to get it to work, otherwise I have to worry about rewriting both label and filename on ROM session requests. |
Session_label constructor now takes a bare string rather than a serialized argument buffer. Replace all instances of previous constructor with 'label_from_args' function. Issue genodelabs#1787
Conveying the ROM filename as the final label element simplifies routing policy and session construction. Fixes genodelabs#1787
|
@nfeske I've split the compound label constructor to a prefixed_label function as you suggested. |
New function 'prefixed_label' for creating compound labels. Issue genodelabs#1787 fixup fixup
Fix namespacing at rom_prefetcher. Issue genodelabs#1787
Session_label constructor now takes a bare string rather than a serialized argument buffer. Replace all instances of previous constructor with 'label_from_args' function. Issue #1787
New function 'prefixed_label' for creating compound labels. Issue #1787
Prevent possible integer underflow in 'Session_label::last_element'. Issue #1787
Fix potential out-of-bounds access in prefixed_label Issue #1787
…ameters) This change belongs to a different commit. Issue #1787
Fix namespacing at rom_prefetcher. Issue #1787
|
I also added the second part of the change to staging, with the only reservation being the part I annotated inline. @ehmry I would very much appreciate a fixup regarding this part. I nevertheless put it to staging to see what the buildbot/autopilot is saying. |
Fixup ROM label rewriting policy. Issue genodelabs#1787
|
I've cleanup the ROM label rewriting at fca6e11 |
Fixup ROM label rewriting policy (again); Issue genodelabs#1787
Fixup ROM label rewriting policy. Issue #1787
Fixup ROM label rewriting policy (again); Issue #1787
|
Thank you for reworking that part. I added both fixups to staging. |
|
Yes, I can think of other situations where |
|
The never-ending story goes on. ;-) I added a few more fixups to staging. Apart from the usual fixes of the automated tests, I changed the new API functions to avoid the use of pointers. This way, we can successively replace the occurrences of |
Session_label constructor now takes a bare string rather than a serialized argument buffer. Replace all instances of previous constructor with 'label_from_args' function. Issue #1787
|
@ehmry Now that this change entered master, it might be a good time to adapt the components of genode-world to it. |
Session_label prefix(a, b) shall not return ' -> b' or 'a -> '. Issues genodelabs#1787
|
I have a small nitpick for the prefix function, d499b29 |
Use ((!x) || y == "") rather than (!(x && y != "")). Issue genodelabs#1787
|
Swapped the booleans around. |
Session_label prefix(a, b) shall not return ' -> b' or 'a -> '. Issue #1787
Session_label prefix(a, b) shall not return ' -> b' or 'a -> '. Issue #1787
As mentioned in #1766, the filename argument can be removed and the rightmost session label element treated as the name of the ROM. By the same logic the root argument on File_system sessions could go the same way, such as
init -> child -> /.The text was updated successfully, but these errors were encountered: