Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

We use GNUTls >= 3.6.7 so this is always true.

Change-Id: Ifcc996a95faf056a3c0eed35cffc17becddeeed3

Change-Id: I157118c875868aa1cae7d0b0ea02644175d33b70

The check in this loop is useless as we already have the transport
that we are searching for. Moreover, locking sipConnsMtx_ from
a pjsip operation (here sipvoiplink) may cause a deadlock.

Change-Id: Ic263d69a3f5518857a93a4a4576a4a63b7fda575

Change-Id: Ief11062f573e331d497aa267a1cd8723a757b113

Multiple changes are included in this patch:
+ Remove pendingConversationFetchMtx_ and use conversationsMtx_
instead to avoid any potential weird state
+ Remove RepeatedTask causing two fetch operations to happen
+ Remove bugguy map access causing a fetch to happen even if
previously removed

A test for the problematic scenario is added to avoid any future
regression.

Change-Id: I9b5f93fc11f053a0552e5f34711130d5a5c22c05

Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

+ Like for windows, add write permissions if possible
+ Rewrite removeAll() condition because it returns an integer.

Change-Id: Id5eb413ed8a6a6345d9208646cdfb13f5c5810bb

Alireza Toghyiani authored 1 year ago and Kateryna Kostiuk committed 1 year ago

Gitlab: jami-client-ios#275

Change-Id: I34c2bd8e34d5c4932dcc2e2555659419c88b6202

When the LIBJAMI_FLAG_NO_AUTOSYNC flag is used, it is still
necessary to call startSync in order to receive invitations.
This patch applies the LIBJAMI_FLAG_NO_AUTOSYNC flag to startSync
callbacks.

jami-client-ios#280

Change-Id: Ieb8a1fb8542c4caf975e13d5a3210bdbc1099a55

GitLab: #855
Change-Id: Iaf9819d082b3b3e43291eb58f0ca59f168971542

Pierre Nicolas authored 1 year ago and Adrien Béraud committed 1 year ago

Allows to fix a bug on android where updatedpreference signal was not emitted

GitLab: jami-android#1271
Change-Id: I015269da19acb18dc0cdfb8310cb12b73aeb585f

Kateryna Kostiuk authored 2 years ago and Kateryna Kostiuk committed 1 year ago

On iOS, both Jami.app and notifications extension could modify
conversations and requests files. This patch adds API to reload
conversations and requests so Jami.app and notifications extension
could have up to date information.

Change-Id: I5873bc69c329bd6c78e7f4d60a01c9471e19bcd1

Else, destruction of the object can occurs during a async task
causing weird crashes.

Change-Id: Ia5f7e8f7324a8e8708c94890627d8576538289d3

+ Use JAMI_ERROR/JAMI_WARNING/JAMI_DEBUG/JAMI_LOG
+ Add a new error signal if an error occurs while comitting

Change-Id: Ic0ec594911bb7d29f6419f9ec1cba060c973d0d6

Change-Id: Ic7044e0aaef48403ab879750f43da202bbee21e0

This commit uses latest gmp version since it fixes
build for arm64 for macOS with assembly enabled.

It does not fix the build for iPhone, so the
option --disable-assembly should still be used when
building for iOS.

Change-Id: I045e0372c36367f22f8b8a95aaac987e559c6d02

GitLab: #854
Change-Id: If260bd73205d579cbe7c963bf3497bff202c1b55

Change-Id: Ic23ace5dbc183781d0020db647a5ac6e77e376fc

To build for arm64 on macOS, Darwin version 20 or higher
is required.

Change-Id: I6da0b78e64eee2b0b3b6169be58fd261ac3085be

Change-Id: I86622eb65679a2c953804fc87344d1733e5cb7b6

Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

Change-Id: I267631faf3bdd24164553de669440c0e8d104638

Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

isCancelled_ wasn't used and cancelOperations should wake up
IceTransport::send to avoid any potential locks.

Change-Id: Ic5d412b5f0e8328f7c64d28dcab5710ed948ccd8

Change-Id: I4d5b2476e22ae9079eabb12012a48d75e308d66d

Change-Id: Ief19e95693560eb1db50a90a446106e201860680

Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

Change-Id: I600651eb9ac15fc081768e14c3bb95c18f44ede5

Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

+ add callMutex_ around rtpStreams_ access to avoid use after free.
+ retrigger setRotation, because it's called from pjsip and we
need to avoid any potential wrong lock order.
+ Fix ut_conference by protecting pInfos_ checks

Change-Id: Ie423c06b1d1f2ef6532afd3144d0667b8f709498

Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

Sync retriggers on ioPool, so it may be invalidated at some point.

Change-Id: Ic2a99d923ec42ce41a396a6d094721d561f44ad5

Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

requests_ is passed as a shared_ptr in verifyCertificateWrapper.
If the session is destroyed, the request will not be stopped and
will not return any result. So we need to cancel it.

Change-Id: If631a3c5409aea05136278d4ab7b1e2d0ae7c3cb

shutdown is not immediate, so the assert was failing. Waiting a bit
that the shutdown finishes fix the test.

Change-Id: Iee88eecd6a78f9115ac057b6321fb032d5feb911

Kateryna Kostiuk authored 1 year ago and Kateryna Kostiuk committed 1 year ago

When an account is registered from the notification extension,
it only needs to fetch commits for one contact and not make
connections with other contacts, since the extension has limits
for time and memory.

It also do not need bootstrap conversation.

Change-Id: Ia5815dfc47176989c8df10766613f063d3211bee

Change-Id: I9c5d79c7e5e66e159990a888964aaa46065359f8

* lock connectCbsMtx_ only once when reading from and writing to pendingCbs_
* only lookup deviceId in pendingCbs_ once, instead of at least twice
* removed not working MAX_TENTATIVE mechanism

Change-Id: Ic4027dcb2f5a1a32e8f73e0f38c0f53451eabf19

Change-Id: I5ab8b5bf1de8ba0dd8b51aa014418fe2d30e3369

Change-Id: Icc7de5a119748b0b608245d7e66139cf330708f8

Sébastien Blin authored 2 years ago and Adrien Béraud committed 1 year ago

This allow to keep different status between accounts, remove
certificates when the account is removed. And this avoid
to use the same certificate store/trust store shared across
accounts.

Change-Id: Iba5df5c0768cefe7a861cab689ffb8ff770d8475

Sébastien Blin authored 2 years ago and Adrien Béraud committed 1 year ago

removeUnusedConnection can happen and will lock infosMtx_ then cbMtx_
and at the same moment onTlsNegotiationDone can happen and lock
cbMtx_ then infosMtx_ causing a potential double lock.
This patch avoid this case by avoiding the infosMtx_ to be locked
if destroying.

Change-Id: If21d35757e9dbd667ef3e20e8c13025479494333

Gitlab: #1091
Change-Id: I0cd0b567045c7a4408be81fc433ac9a40f1031a3

Andreas Traczyk authored 1 year ago and Adrien Béraud committed 1 year ago

Change-Id: I0d8c9172b7d276afb4509849406f227b61d36f11

Change-Id: Ifd69193316a4bebe373968675721916505e10b65

fmt::format line was incorrect due to the brackets.
Also add a single test doing a POST request to avoid any regression

Change-Id: If6aa44885d5c09ed4e931b1437d6294326e8d441

Sébastien Blin authored 1 year ago and Adrien Béraud committed 1 year ago

Change-Id: Ie1e33d9638f2ab616764c8c29e877bdbf3b7f1ea