Don't just silently ignore invalid user data value, this is not supposed
to happen.
Also use "userdata" name for the same parameter of all callbacks
consistently.
As this flag is tested in the worker thread only when it owns the mutex,
set it only after acquiring the mutex too to avoid any possibility of a
data race.
Call SetState(State_Active) before signalling the worker thread, as
otherwise it would be possible for it to wake up and set its state to
something else before it was reset to "active" from the main thread.
This fixed another TSAN error.
As in 31a441e814 (Use separate test cases for wxWebRequest auth tests,
2021-01-10), there is no real reason to use sections here as we don't
reuse anything between them and using separate tests makes it easier to
run individual tests and, especially, combinations of them.
This commit is best viewed ignoring whitespace-only changes.
Mutexes must not be destroyed while locked and thread sanitizer
correctly complains about this, so ensure that we do unlock the mutex
before the worked thread terminates and the object is destroyed.
Also use critical section instead of a mutex, as this is more efficient
under MSW.
Main purpose of this commit is to make it clear that this mutex/critical
section is only used together with the data from the same struct.
No real changes.
Don't use system libcurl if --disabled-sys-libs was used, but warn about
it as this could be unexpected.
We could consider including libcurl as submodule, and use the built-in
version as a fallback, but it's a relatively big repository (100MiB+),
so it's not clear if it would be a good idea to do it.
This allows to easily run both auth tests at once without all the other
tests and it's not like we reuse much (or actually anything) by having
them as sections in the same test case anyhow.
This commit is best viewed ignoring whitespace-only changes.
Just sort the features in the same section in alphabetical order to make
it easier to find the given feature in the list (or confirm that it is
not present).
No real changes.
Use tool position as offset from group position and not vice versa, both
because this was already done like this in the other places and because
this makes more sense: the tool is inside the group, so its position is
relative to it.
No real changes.
Don't update wxRibbonToolBar and wxRibbonButtonBar tooltips each time
the mouse was moved, this was inconsistent with the standard tooltips
behaviour and resulted in flicker.
Also don't show tooltips at all over a dropdown, as they partially
covered and, again, such behaviour is very non-standard.
Closes https://github.com/wxWidgets/wxWidgets/pull/2162
This allows to write the code using this class without peppering it with
wxUSE_SECRETSTORE checks that would otherwise be necessary to support
Unix builds on system without libsecret.
No real changes.
No real changes, just try to avoid over long lines.
Also use early returns in case of WinHTTP functions failures everywhere
for consistency.
This commit is best viewed ignoring whitespace-only changes.
This allows to avoid using temporary variables just to be able to pass a
pointer to them to WinHttpSetOption().
No real changes, just a simplification.
This doesn't seem to be needed, our callback has the correct signature.
If it's required for some non-MSVC compilers (e.g. MinGW with old SDK),
it would be better to use the cast only conditionally to at least keep
the MSVC build type-safe.
It seems better to rely on the well-tested WinHTTP URL parsing functions
rather than on our own wxURI. It should also allow to support any new
URI schemas if support for them is ever added to WinHTTP.
No real changes, just avoid using raw numbers, which is not very
readable, notably for HTTP_STATUS_PROXY_AUTH_REQ which is not really a
well-known value.
Check that current state is State_Idle in wxWebRequest itself only once
instead of doing it in 2 (out of 3) wxWebRequestImpl implementations.
Also assert if this is not the case instead of silently doing nothing
which would surely be more difficult to debug.
Don't duplicate calls to ::FormatMessage(), which is difficult to use
correctly, in wxCrashReport and wxWebRequestWinHTTP, but just reuse the
same code that was already present in wxSysErrorMsgStr() after
refactoring it into a reusable function allowing to specify the module
name to use for the error code lookup (before falling back to
interpreting it as system error code).
This fixes not trimming the trailing "\r\n" from the string in the other
places (wxWinHTTPErrorToString() had code to do it, but it was wrong,
while wxCrashContext::GetExceptionString() didn't do it at all) and
avoids duplication.
Drawing a circular point into a square pixel does not completely fill the pixel
with the new color, resulting in an alpha-blended appearance. Instead, draw a
square into the pixel. As a bonus, this is much faster, at least with Cairo.
See #19037
Instead of doing it the other way round, as it has several advantages:
1. wxSysErrorMsgStr() is not limited by the static buffer size any
longer, i.e. doesn't truncate errors even longer than 1KiB.
2. Code is much simpler and more obviously correct.
3. We avoid an extra and absolutely unnecessary buffer copy.
No real changes.
Just micro cleanup: there doesn't seem to be any need to show which
function we're in as FormatMessage() is only called from a single place
anyhow, so shorten and simplify the code.
Rename Init() to Open() as we need this method to return bool to
indicate its success in order to avoid using non-initialized handle
later. Init() is also reserved, by convention, for the common part of
all class ctors in wx code.
Remove m_initialized entirely, it doesn't seem to be obviously better to
cache the failure to create a session than to retry doing it every time
(in fact, it would seem to be worse) and not having it is simpler.
This commit is best viewed ignoring white space.
Remove an explicitly set CMAKE_OSX_DEPLOYMENT_TARGET from the CMake iOS
build in order to use the default value which is now also set to an iOS
instead of macOS version.
The minimal sample doesn't have an implicit deployment target set, while
the library does through CMakeLists.txt in root. This can result in iOS
link errors when not explicitly setting CMAKE_OSX_DEPLOYMENT_TARGET as
the sample will then use a target OS with an architecture that may not
be present in the library with its current target OS version.
Fix by copying the part setting CMAKE_OSX_DEPLOYMENT_TARGET in the
library to the minimal sample as well. Both parts should be kept in sync
but that still seemed preferable to additionally polluting an example
file for CMake usage with a dependency by introducing a common include
file.
Differentiate between the minimal supported version of macOS (10.10)
and iOS (12.0) instead of setting CMAKE_OSX_DEPLOYMENT_TARGET to
(macOS-only) 10.10 for both platforms.
Cancel the request and wait until it actually is cancelled when exiting
the sample.
This is a bit ugly, especially the busy-waiting part, but still better
than potentially crashing.
A cancelled request is not supposed to end up in the "failed" state, but
it did, resulting in showing an error in the webrequest sample after
pressing on the "Cancel" button, which was clearly unwanted.
When cancelling a request it was possible for it to be deleted while it
was still used from another thread.
Fix this and make the code much more obviously correct by simply
"locking" the request until the event generated by it is processed: now
IncRef() and DecRef() calls are always balanced as they're called from
ctor and dtor of StateEventProcessor helper only.
