Notes |
|
|
urldb asserting
"content/urldb.c", line 1964: urldb_match_path: Assertion failed: path[0] == '/'
( 6ad498) pc: 44610 lr: 44b90 sp: 6ad49c ^urldb_find_url()
( 6ad4a8) pc: 44b7c lr: 50d1c sp: 6ad4ac urldb_get_url_data()
( 6ad4d0) pc: 50c14 lr: 3217bc sp: 6ad4d4 ^node_is_visited()
( 6ad508) pc: 3211cc lr: 3220bc sp: 6ad50c ^match_details()
( 6b298c) pc: 321a88 lr: 4a9c1c sp: 6b2990 css_select_style()
( 6b29b8) pc: 4eb64 lr: 86e04 sp: 6b29bc nscss_get_style()
( 6b2b34) pc: 86544 lr: d2a28 sp: 6b2b38 ^convert_xml_to_box()
( 6b2b54) pc: d29e0 lr: be6a4 sp: 6b2b58 schedule_run()
( 6b2e1c) pc: be468 lr: bf750 sp: 6b2e20 ^riscos_poll()
( 6b2fe8) pc: bf0d0 lr: 4a9700 sp: 6b2fec main() |
|
|
|
If you can reproduce this I would be very interested in the new log, most recent CI build (2266) has additional logging in this part of urldb which should let me understand what bad bad urldb is choking on. |
|
|
|
I've just tried that URL again with development versions 2264 and 2266, and there was no crash. However, I had to access the URL directly, not through www.bbc.co.uk/news, as the original page is long gone/ This may mean that the steps to reproduce couldn't be done exactly. |
|
|
|
Thanks to the duplicate 2215 giving the additional debugging hint I needed I can now say this is caused by urldb choking when asked to check the visited status of mailto: urls
In the case of the example on the register it was their "Tips & Corrections" link which is
mailto:corrections@theregister.co.uk?subject=%5Bcorrection%5D%20BBC%20clamps%20down%20on%20ILLICIT%20iPlayer%20watchers&body=re:%20http%3A%2F%2Fwww.theregister.co.uk/2014/11/03/beeb_clamps_down_on_naughty_iplayer_watching/
and for some reason on RISC OS is causing the path assert to be triggered (this does not occur on GTK) |
|
|
|
I reported this one, and haven't come across a simiar crssh since. I use the BBC news site quite a lot, and I think this bug could be closed now. |
|
|
|
I've tried again with RISC OS build 3252, and was again unable to reproduce the problem, so, as the OP suggests, I'm closing it. |
|
|
|
> and for some reason on RISC OS is causing the path assert to be triggered (this does not occur on GTK)
Nothing to do with RISC OS/GTK, but rather the contents of your URLdb.
When we test whether a URL exists in the DB for :visted support we use a bloom filter to speed up the case where its not there. If there is a hash collision then we actually have to do the test fully and hit this assert. |
|
|
|
Confirmed fixed in 3.4 release |
|