Skip to contentSkip to conversation

harfbuzz: CVE-2023-25193 doesn't fix the issue and still exhibits O(n^2) behaviour #kirkstone

PATCH In review DC 3 participants 3 messages Mar 6, 2023

lore
Share & tools
lore

DC

author
#0 lore raw

PATCH 1/1 harfbuzz: CVE-2023-25193 doesn't fix the issue and still exhibits O(n^2) behaviour #kirkstone

The patches submitted to
dunfell -> https://lists.openembedded.org/g/openembedded-core/message/177768
Kirkstone -> https://lists.openembedded.org/g/openembedded-core/message/177767 ( https://lists.openembedded.org/g/openembedded-core/message/177767 )

Still exhibits O(n^2) behaviour upon testing with applications and fonts and in worst cases Harfbuzz is not printing anything. Upon closer examination, looks to me that a wrong fix is pushed for the issue.

The patch [ https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc ] is known to causes erroneous results and regression already reverted by Behdad as a developer of Harfbuzz in the commit [ https://github.com/harfbuzz/harfbuzz/commit/661050b4659ee490dfe622821bc7fde7d1c40510 ].

(ref-> https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc#commitcomment-101326711 )

The new fix was already introduced ( https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc#commitcomment-101335712 ) and needs to be backported properly.

It doesn't look correct to patch Harfbuzz with code already reverted.

#1 lore raw

Yes, will revert that patch and backport the correct one.

Thanks
vivek

Quoted 42 lines · DC

The patches submitted to
dunfell ->
https://lists.openembedded.org/g/openembedded-core/message/177768
Kirkstone ->
https://lists.openembedded.org/g/openembedded-core/message/177767
<https://lists.openembedded.org/g/openembedded-core/message/177767>

Still exhibits O(n^2) behaviour upon testing with applications and fonts and in worst cases Harfbuzz is not printing anything.
Upon closer examination, looks to me that a wrong fix is pushed for the issue.

The patch [
https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc
] is known to causes erroneous results and regression already reverted by Behdad as a developer of Harfbuzz in the commit [
https://github.com/harfbuzz/harfbuzz/commit/661050b4659ee490dfe622821bc7fde7d1c40510
].

(ref->
https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc#commitcomment-101326711
)

The new fix was already introduced (
https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc#commitcomment-101335712)
and needs to be backported properly.

It doesn't look correct to patch Harfbuzz with code already reverted.

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group. View/Reply Online (#178056):
https://lists.openembedded.org/g/openembedded-core/message/178056
Mute This Topic: https://lists.openembedded.org/mt/97419449/7129709
Mute #kirkstone:
https://lists.openembedded.org/g/openembedded-core/mutehashtag/kirkstone
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [ [email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

#2 lore raw
Quoted 2 lines · vkumbhar

Yes, will revert that patch and backport the correct one.

I will remove that patch from the current kirkstone queue and await a V2. It hasn't moved out of my testing branch, so a revert isn't needed.

Steve

Quoted 30 lines · DC
Quoted 8 lines

The patches submitted to
dunfell -> https://lists.openembedded.org/g/openembedded-core/message/177768
Kirkstone -> https://lists.openembedded.org/g/openembedded-core/message/177767

Still exhibits O(n^2) behaviour upon testing with applications and fonts and in worst cases Harfbuzz is not printing anything. Upon closer examination, looks to me that a wrong fix is pushed for the issue.

Quoted 11 lines

The patch [ https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc ] is known to causes erroneous results and regression already reverted by Behdad as a developer of Harfbuzz in the commit [ https://github.com/harfbuzz/harfbuzz/commit/661050b4659ee490dfe622821bc7fde7d1c40510 ].

(ref-> https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc#commitcomment-101326711)

The new fix was already introduced (https://github.com/harfbuzz/harfbuzz/commit/85be877925ddbf34f74a1229f3ca1716bb6170dc#commitcomment-101335712) and needs to be backported properly.

It doesn't look correct to patch Harfbuzz with code already reverted.

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#178057): https://lists.openembedded.org/g/openembedded-core/message/178057
Mute This Topic: https://lists.openembedded.org/mt/97419449/3620601
Mute #kirkstone:https://lists.openembedded.org/g/openembedded-core/mutehashtag/kirkstone Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-