When building with GCC 13, this test no longer builds correctly and errors out
because the overloaded `operator<<` for `BPrivate::BDate` and `BPrivate::BTime`
can not be found.
The underlying cause is that the libstdc++ depended on some template SFINAE
behavior, and that did not work because the `operator<<` should only be looked
up in the `BPrivate` namespace.
It is unclear whether the code for libstdc++ changed, or whether this never was
supposed to work. GCC bug #51577 seems to imply that GCC before version 12
errornously allowed the lookup of names in the global namespace. It could be
that we always relied on this behavior:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51577
Change-Id: Ia2f2306a2e97d5f19dc8c4df90a8491f22ef0bcd
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6874
Reviewed-by: Adrien Destugues <[email protected]>
Tested-by: Commit checker robot <[email protected]>
The main motivator is the update to GCC 13. As part of this, some of the other
dependencies have been updated as well.
Newly introduced:
- gawk on x68_64 (new dependency of texinfo)
- libjx (new dependency of haikuwebkit)
- openexr30 (new indirect dependency of haikuwebkit)
- brotli (new indirect dependency of haikuwebkit)
On x86_gcc2, some packages have switched to the modern GCC version:
- diffutils
- findutils
- libpsl
- tcpdump
Change-Id: Ic617b5b4af9eb34c0d28259a3c0ddbcc33f98a5d
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6772
Reviewed-by: Adrien Destugues <[email protected]>
GCC 13 warns that using the `begin()` method on a container is marked as
`[[nodiscard]]`. This makes sure that this warning is not treated as an error.
The code actually looks like it does not do much if anything of use. As a
follow-up ticket #18478 is raised.
Change-Id: Ie149f7e02cacda2bcd0be59617583605d5905747
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6655
Reviewed-by: Niels Sascha Reedijk <[email protected]>
Tested-by: Commit checker robot <[email protected]>
GCC 13 warns that the item could be initialized. From looking at the code, it
could very well be that the log message contained garbage. At least clear the
data before sending it out.
Change-Id: Id02bf314d0c02f852748bdac9cef010ebf073994
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6654
Tested-by: Commit checker robot <[email protected]>
Reviewed-by: waddlesplash <[email protected]>
The module includes the C-style header file, thus all members are added in the
global namespace.
Change-Id: I5e82c8520b0495e3c437c43cfeea9cc454d39482
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6653
Reviewed-by: waddlesplash <[email protected]>
GCC 13 warns that this member could be unitialized. Now it is initialized
explicitly before it is used anywhere else.
Change-Id: Ia8d5e95cc247921c6ee240a106a80d73c54caee4
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6652
Reviewed-by: waddlesplash <[email protected]>
The C++ standard in rule 3.3.7[basic.scope.class] determines how the compiler
should handle cases where a member shares the name of a type. It is up to the
compiler whether this requires a diagnostic warning. GCC 13 emits a warning
(changes-meaning). This fixes the issuing of this warning.
Change-Id: I9c58c0dd6298055959154f78d47ad9088e8225dc
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6651
Reviewed-by: waddlesplash <[email protected]>
GCC 13 warns that the a pointer is taken from a member that according to the
rules is not yet initialized. Fix the declaration and order of initialization
to prevent us depending on undefined behavior.
Change-Id: Iab4644c33a3a286e083450da0b3575acfe284f29
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6650
Reviewed-by: waddlesplash <[email protected]>
Just in case, adjust the buffer to format an int32 to a string as the maximum
length that it could be.
Change-Id: I00aede3f684211f9c860455063244422be1d0ee9
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6649
Reviewed-by: Adrien Destugues <[email protected]>
With GCC13 the current code and its use of the `realloc` function triggers a
potential use after free warning. The code is most likely correct, but there
may be a bug in GCC13's use after free checking, or this is one of the cases
where it is not possible to accurately detect if there is really a chance that
it is used after free.
This rewrites the code in such a way that the warning is no longer triggered.
No functional change intended.
Change-Id: I85e145de2128d4b12a9b3016de33d9facaf0d758
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6648
Reviewed-by: waddlesplash <[email protected]>
Reviewed-by: Axel Dörfler <[email protected]>
In C++20 `requires` is a reserved keyword. GCC13 warns against its use. This
change removes the use of the keyword in our own code, and ignores it in the
external 'libsolv' code.
Change-Id: I3319d20c0a3fdad89630683cdff5ea3b17f04e99
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6647
Reviewed-by: waddlesplash <[email protected]>
GCC 13 is more strict about what C++ features are available in a freestanding
(= build without standard library) build. The `stdlib.h` and `math.h` headers
are redefined as part of libstdc++. If the object is built as freestanding, it
should be assumed that none of the standard library functions are available.
The solution in Haiku has been to add part of the standard library into the
kernel shared object. The method of exposing them has been to allow the
use of `stdlib.h` and `math.h` in the kernel and in kernel add-ons.
This change allows that approach to continue. What it does, is that it defines
specific headers which will be picked up when a module is built using private
kernel headers. When building for the kernel (or for the boot module), it will
bypass GCC 13's libstdc++ default behaviour of not including the POSIX/platform
header with all the function definitions.
An alternative to be considered for the longer term is to specifically define
the parts of the C/C++ standard library that is available in the kernel in
these headers, or create custom headers when building kernel modules (which is
how Linux approaches it).
Change-Id: Icab4614f642219fa77732b02401570708ee9a963
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6645
Reviewed-by: Adrien Destugues <[email protected]>
The query bin command is already added earlier in the file, and does not
actually require to be linked against localestub. Removing the duplicate, as
it caused all the include directives to gcc to be duplicated, which caused
trouble in gcc2 when using #include_next directives.
Similarly, urlwrapper.cpp is a duplicate entry and does not actually require
libbnetapi.so.
The clockconfig tool has moved to the launch daemon in c16361c4.
Change-Id: I266ada433448a96bee8de325133a548f0985da59
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6661
Reviewed-by: Adrien Destugues <[email protected]>
GCC 13 warns that the use of auto_ptr on an array can lead to undefined
behavior, because auto_ptr will use `delete` instead of `delete[]`.
Replaced the use with `std::vector` instead. Since the `data()` member was only
introduced in C++11, the solution uses the [] operator instead to get the
address of the beginning of the array.
Change-Id: Ib457580eedb03338a454fb96b27602c3830c1634
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6644
Reviewed-by: Adrien Destugues <[email protected]>
Tested-by: Commit checker robot <[email protected]>
GCC's overflow checker gives a warning about the potential that the atomic
functions called in teams.cpp and threads.cpp will write to a null-pointer.
In this case, it is safe to assume that the values will never be null,
therefore these warnings can be safely ignored.
Relates to/fixes #17734
Change-Id: I777ff96f2812ed7d3ba57a46cad89ef6994cb08f
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6643
Tested-by: Commit checker robot <[email protected]>
Reviewed-by: Adrien Destugues <[email protected]>
GCC 13 is more strict about what C++ features are available in a freestanding
(= build without standard library) build. The container types are not
available.
In this case the device manager does not use the std::set container.
Change-Id: Iaa2d71b40c9247fac8ccf2a0033c9fc288550dc7
Reviewed-on: https://review.haiku-os.org/c/haiku/+/6642
Reviewed-by: waddlesplash <[email protected]>
Updated so that it can build Haiku R1 Beta 4. The intention is to use this to
create repeatable builds for Rust builds. The script has been tweaked to use
the latest stable debian as base image, to include the (now-required)
libzstd-devel package, and the g++-multilib package to support building 32 bit
in a 64 bit host. It also moves to python 3.x.
Additionally, the image now supports specifying custom git tags/branches for
the buildtools and Haiku itself, so that specific hrevs or branches can be
built
Change-Id: Ia43089ff7271256fdb6603bd4400ee3b9228ea1b
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5965
Tested-by: Commit checker robot <[email protected]>
Reviewed-by: Adrien Destugues <[email protected]>
The API allows a user to end parsing responses earlier if the response has a
client (4xx) or server (5xx) error code. The end user should still be notified
of this result, so that they can act.
Change-Id: I11f2ab296658f767b5b34e10badb54de27aab494
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5910
Tested-by: Commit checker robot <[email protected]>
Reviewed-by: Adrien Destugues <[email protected]>
Reviewed-by: waddlesplash <[email protected]>
The overall design does not deviate much from my proof of concept [2] and that still makes a good read to
understanding the overall architecture. If you want to get a sense of how it is built up, the API comes with
full doxygen documentation for the public API [3], and I have also done a PoC change for HaikuDepot which is
useful as an illustration on what the impact for the user of the new library is. [4] There is also a test suite
that may give some insight into the day to day ergonomics of the API [5].
The current state is that I am fairly confident that many HTTP requests will actually work, but I do expect
rough edges with a protocol with this many diverse implementations. There is also a list of features yet to be
implemented on Trac [6]. Additionally, I still want/need to do performance testing.
The goal of merging the kit right now is to start making it available for more uses, and through that also give
a chance to shape its future. There are also some design decisions that need review, most notably I expect some
discussion around the uses of C++ 17 idioms (like std::optional and std::string_view) and around the use of
exceptions for error handling.
The impact of merging right now should be near zero: the netservices2 kit lives in its own header space, and
builds into its own static library (libnetservices2.a). It is not yet used in any of the apps in our
repository.
The branch does remove the deprecated services kit from the libnetapi.so library, though it leaves
libnetservices.a intact. After our previous announcement to remove it after beta 3, this should be expected.
[2] https://github.com/nielx/haiku-netservices-rfc/tree/exceptions
[3] https://git.haiku-os.org/haiku/tree/docs/user/netservices?h=dev/netservices
[4] https://review.haiku-os.org/c/haiku/+/5692
[5] https://git.haiku-os.org/haiku/tree/src/tests/kits/net/netservices2?h=dev/netservices
[6] https://dev.haiku-os.org/wiki/Development/NetServices2
Change-Id: I5d0b7e2619699f39a2506588417b57391f0f5cc2
This commit formats all the netservices2 code with the `haiku-format` tool from
https://github.com/owenca/haiku-format (commit aa7408e), with the following
customizations:
* SpaceBeforeRangeBasedForLoopColon is set to false
* Braces before a catch block are not wrapped
* Most headers, except for ExclusiveBorrow.h, have been manually reformatted
to adhere to Haiku's header format (issue #19 in the repository)
Change-Id: I693c4515cf26402e48f35d1213ab6d5fcf14bd1e
When parsing a raw HTTP body without a known size, whether or not a request is complete depends on the
end of the connection. Make sure that the parser updates to complete when all the bytes in the buffer have
been read and the connection closed.
Change-Id: I6f055b43ffe4a44da65da85c19b71304d804c800
This change refactors various parts of the HTTP parsing. The HttpParser now
tracks what part of the message needs to be received next, and throws an error
if the object is used in the wrong way (when the caller requests to parse the
wrong part of he message).
The metadata about the transmission is now also saved in te parser. There is
partial work in there to start exposing the 'bytes written', which in
compressed streams is expected to differ from the bytes read. This is not
used yet.
This also simplifies the state tracking done by BHttpSession::Request.
Change-Id: I8532c6a5c8776456ea8bbccd6df7a44bac92b60d
When change 2966 was merged as hrev56269, it was decided to first introduce
this in libshared instead of libbe.so. The test was moved accordingly, but
not all the code in the unittests was properly updated. This change completes
it.
Note that the test currently crashes, so it needs more investigation whether
that is a test issue or a BMemoryRingIO issue. Fixing this is beyond the
scope of this change.
Change-Id: I0f61d1dd013da904b3e4545db11fefa9da2da0a2
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5676
Tested-by: Commit checker robot <[email protected]>
Reviewed-by: Jérôme Duval <[email protected]>
This smart pointer is designed to help with putting some explicitness and
safety around the case where someone will use their own object that implements
the BDataIO interface to store the body of a network request. By default,
BDataIO objects do not require or enforce thread safety. Since accessing these
unsynchronized objects between two threads is undefined behavior, it should be
explicitly discouraged.
The BExclusiveBorrow/BBorrow smart pointer helper helps solve that by enforcing
the limitations on using an unsynchronized object in two threads. When used
correctly, there is a runtime check on incorrect use by the developer. This
should help write better code.
The design is based on shared_ptr, including having an admin block akin the
control block, that manages the internal object. This type-erased admin block
has the advantage that it allows the owner to have a different type than the
borrower. It also handles cases where the lifetime of the borrower is longer
than the owner: the borrower can continue to use the object until they want to
return it, after which it will be cleaned up. This will make it possible to do
some fire and forget pattern in the network services kit, where someone may
just wants to create a file and borrow it to the network request, and care
about further processing the file in the future.
Change-Id: Ie9b7e7472c868b60f663b4db4fa449d421e447eb
These allow for additional messages about the progress/status of the request.
For now, the messages are sent unconditionally, though it may be reasonable to
in the future switch to sending them only when enabled.
Change-Id: Ic45a0055037af02c689494fa5ce0acd03592ad7e
This fixes a bug in the HttpAuthTest, and in general, moves responsibility of determining
the shape and size of the content to the right place.
Still to do is to fix the case where there really is a variable length content. This will
now probably error out as a connection closed error.
Change-Id: I13c20028e834cc9af8c7bc1d7d7613bf60838e64
This change allows the user to control how many concurrent request can be done
per session. This is going to be helpful to running the tests as well; they were
all fired up in parallel, which sometimes leads to our test server refusing a
connection.
Change-Id: I4f1f40b28b8e0199ea5589b36cd8d00ecd849a63
The integration PostTest has a basic test that the expected messages are sent and
have the expected data fields. The gist is documented in book.dox.
To do are the messages around SSL. However, that functionality is also not
implemented yet, so there is nothing to send.
Change-Id: Ib8f36ed32f9854d643d8256338b71af7067059f0
In the old service kit, the Accept header was automatically set. However, it is
not a required header. This removes the default value and the protection against
setting it manually.
Required to make HaikuDepot work.
Change-Id: Ic589bfc5829db25915b67f2a13b6c2436c053e25
The previously implemented behavior of strptime() allowed for partial matches
of strings. For example, if the format string was "%a, %d %b %Y %H:%M:%S", it
would successfully parse "Sun, 07 Dec 2003" as input. This is inconsistent with
the standardization in POSIX.1-2001 (and later).
This change makes strptime() return an error if there is no more data in the
buffer and not the entire format string is parsed.
Change-Id: If066c49fb7fc094f8ccd56703cd01903a0e40cb3
Reviewed-on: https://review.haiku-os.org/c/haiku/+/5298
Tested-by: Commit checker robot <[email protected]>
Reviewed-by: Jérôme Duval <[email protected]>
These utilities convert timestamp strings that are formatted according to
the HTTP RFC into BDateTime objects, and vice versa.
Change-Id: Ia2498944fb63d09233839f19d08f15d82a0a9685
The user of the API can set whether redirects should be followed, and if so,
how many. This is part of the BHttpRequest API. The BHttpSession then follows
those instructions, and executes the maximum number of redirects the user
would like to follow.
As part of this commit, the BHttpStatusClass and BHttpStatusCodes helper enums
have been added, to give a friendlier access to HTTP status codes and status
classes.
Change-Id: Ic8c9e3fda158e2cce549c8f1d360951f7ac83311
These particular responses will not have a body. This is now handled by the
BHttpSession object. There is also a minor fix in here that prevents a crash
when multiple requests are handled by the DataThread at the same time, and not
all of the requests have events.
Change-Id: I7f47d8b3cd8491c8193275be4b3fc1080780fa20
This initial implementation contains a lot of extra debuggin output that will
be removed in future changes. It is now possible to run a simple GET request
to the test server with gzip encoding.
Change-Id: I2c402e5cf80b94b366563888222a891a1b094b7a
This change also drops the principle that fields with the same keys would be
grouped together. This was initially inspired by Boost::Beast, but it means a
lot of extra copying of data when adding/organizing the list, as well as
inefficient querying on each add. Now that the design choice is to fully go
for the raw string as underlying data storage, that choice is not necessary.
In the future it may be able to emulate the grouping or retrieving of lists
of values in the API, rather than as a fundamental principle of the data
storage.
Change-Id: I2667cfa38eb3b7b75393ee71fb038231a40b4193
BHttpSession::Execute() moves the request into the session, and returns a future BHttpResponse
object. Currently implemented are resolving the hostname, and opening the connection.
There is some scaffolding for the actual data transfer.
Change-Id: I5a8a7a7f8680036b91cdba4beee140bbed6bfd5a
This is a generic error type that can be used by multiple protocols to describe errors that can
occur while processing a request. The error type supports adding an additional error code in cases
where there is an underlying system error.
The type will be used to describe errors that occur while processing requests by BHttpSession, and
it is generally going to be thrown by the receiving BHttpResult.
Change-Id: I76c0bbaedd38df8cfb79159c4beae2fbf1350aab
Incomplete class, but will provide the basis to start working on the internals of the BHttpSession.
Change-Id: I3ca14b7bd823fc1b4a5a32f5784592d214c4e9a7
Objects of this class describe a HTTP request. It contains several convenience
functions that will allow a user to describe the properties of the request.
More options to be added later.
Change-Id: If6a00d26808c5ed4b121cb36dc75a2a1cc449f95
This class provides defaults and performs basic validation for HTTP Methods as
defined by the standard. They will be used in conjunction with a future
BHttpRequest class.
Change-Id: If69a7ec186d9d1165e8efe5ab5df50d5a089208d
HTTP messages (requests and responses) have a header section that can contain
HTTP headers. These headers consist of name, value pairs. This class can be
used to query the headers on a response, and build a list of headers for a
request.
The internal implementation is designed around two different methods of storing
the underlying data. For HTTP requests, the name, value pairs are stored as
owned BString objects. For responses, the assumption is that there is a byte
buffer that contains the data and that has the same lifetime as the BHttpFields
object. The name, value pairs will then be stored as std::string_view to the
underlying buffer.
Still to do is:
- The method to convert a BHttpFields list into a string buffer to transmit.
- The method to parse a string buffer and turn it into a BHttpFields object.
Change-Id: I4819db100aa1671aa7403675216a4c85fd221da7
<strstream> has been deprecated since C++98. Both GCC 2.95.3 and GCC 8.3 have
the successor <sstream> available, so use that one instead.
Change-Id: Ifefe686974864c2fd1d4a9d083294b1edb436fbd
The pre-allocate functionality was added to Haiku in hrev54704. It needs
support at the file system level though, and it is yet to be implemented by
BFS. Some applications (and glibc!) implement a fallback mechanism using
ftruncate(), including the LLVM tools, but they go to this fallback
mechanism when it is clear that the operation is not supported. In particular
they look for EOPNOTSUPP.
The current implementation returns B_UNSUPPORTED from the vfs layer when a
file system does not implement the feature. This error code this not map to
a POSIX error. This change converts it to EOPNOTSUPP.
Change-Id: Ief382b0f4d462dfedf84c731f68f69731de4498c
Reviewed-on: https://review.haiku-os.org/c/haiku/+/4492
Reviewed-by: Adrien Destugues <[email protected]>
Reviewed-by: Jérôme Duval <[email protected]>
Tested-by: Commit checker robot <[email protected]>
This removes the use of the destructor in the move assignment operator, as it
may rely on undefined behaviour from the compiler. Additionally, some duplicate
logic to dereference and free a shared string has been unified under
_ReleasePrivateData().
Change-Id: Ie9f51d598c734f83cd0fba49b651315c6e9c8aac
Reviewed-on: https://review.haiku-os.org/c/haiku/+/4440
Reviewed-by: waddlesplash <[email protected]>
Tested-by: Commit checker robot <[email protected]>
This implements the "rule of 5" for this type. While the copy operation for
BString was already using shallow copies of the underlying data, this change
further optimizes moving the data from one object to another.
While it is not the intention to implement move semantics to all types in the
legacy Haiku/Be kits, data types like BString are good candidates, because move
operations are often useful when working with data within an application.
In this implementation, the internal data of the string object will be set to
NULL, thus leaving an empty string.
Change-Id: I16bf9424f9b17f622b0b57659b80628e18760288
Reviewed-on: https://review.haiku-os.org/c/haiku/+/4428
Reviewed-by: Jérôme Duval <[email protected]>
Previously, hrev55011 introduced localization for Cortex, which then
prompted to close#7530. It has then been reopened, when it turns out
that the translations for the container application (RouteApp) did not
work, whereas the individual add-ons/modules were translated.
The cause is that by default BCatalog looks up the translations based on
the subtype part of the signature. This is x-vnd.Cortex.Route (without
the application/ supertype). This change will place the translations in
the right place of the file system.
The add-ons were never affected, since they BCatalog is explicitly told
to find the translations for the entire signature, like:
static BCatalog sCatalog("application/x-vnd.Cortex.InfoView");
Even so, it was chosen to omit the `application` supertype from the
signature for the shared code as well.
This should fix#7530 for good.
Change-Id: Iff18fabef7aba68602e49db1e98cfed2f486f545
Reviewed-on: https://review.haiku-os.org/c/haiku/+/4091
Reviewed-by: Niels Sascha Reedijk <[email protected]>
Tested-by: Commit checker robot <[email protected]>
