Age | Commit message (Collapse) | Author |
|
Since we updated ABE's default gdb.conf to point to gdb-8.2, a new
binary is build: gdb-add-index. This makes this test fail.
This patch fixes it by adding the new name in the expected list.
Do not bother with handling legacy binary toolchains lacking this file
(some validation jobs run this test against the 'latest' release,
which will soon have this file anyway).
Once our next binary releases are out, it would be difficult to handle
this case nicely, unless we hardcode release versions in the tests
(because we'll have offcial binary releases with and without this new
binary, not only "regtest" ones).
Change-Id: I35f1b0d01c46f71aa4e13693040a87d56db7fe72
|
|
It seems the executable name we generate is not "unique enough", and
can lead to race conditions when the board access is slow and two
validations run in parallel. That's probably because we use
containers, and thus the process ID is always the same between two
runs of the same tests. Try to workaround that by using hostname and
date.
Change-Id: Idfd58feeeaed02a2869a18777b5b6a30e0b53dec
|
|
Check that gdbserver is not statically linked.
For this, we check that the requested interpreter is the right one,
and that some dynamic libraries are needed.
Indeed, linking with --static would still generate a binary with a
requested interpreter because it is created using
-Wl,--dynamic-list=${srcdir}/proc-service.list.
The default interpreter can be wrong (eg on AArch32), but even if
right, the generated executable is not functional.
Change-Id: I64bb686f02120c30e40bdfbb5c9a4395f7df17e6
|
|
gdbserver must appear at the end of the reference list, since we
explore the sysroot path after bin/. This makes a difference on mingw
toolchains, where some support libraries used to be listed after
gdbserver.
Change-Id: I6148ef2ec67ca7ccee03ecc81ce5a440e9a20274
|
|
We have now moved gdbserver into the sysroot, so don't expect to find
it in the main toolchain bin/ dir.
Change-Id: Ic5e8d778d61d0767c6758fdb0a74c628700d26cd
|
|
We do not ship it anymore, so don't expect it.
We don't complain if present, because we could be checking an old
toolchain.
Change-Id: Ie6622c3e47e3f70c907b8fda9e2dbea6e407ea0e
|
|
Changes:
* Use HOST instead of TOOLCHAIN to detect if we are in mingw environment.
* Pass HOST to corresponding sub-makefiles in this repo.
* Use *mingw* pattern instead of *mingw32 (in bug-1925) to make code the same everywhere.
* Update README according to changes.
Change-Id: I8085f2049a2dddd8e748ed99ae02647bf0e3eb30
|
|
Change-Id: I4cf89e27639421c40164f76d973c763dc0ab7811
|
|
Add a regression test using the test case provided in
https://bugs.linaro.org/show_bug.cgi?id=2915
Change-Id: I278fd08bd94bc283294648fb7d7ea4e954b5a3b3
|
|
Improve coverage for this test by enabling it for arm* toolchains.
Change-Id: I2ff2b4dc0a5621fc775c0fab3806ca90ca14024d
|
|
Replace the preprocessed contents of stdlib.h with
"#include <stdlib.h>" to avoid a mismatch which occurs when
glibc's header files which are pulled in by arm_neon.h are
inconsistent with the preprocessed version inside
loopfiltersimpleverticaledge_neon.preprocessed.c
Change-Id: If0d311e5f74c9ba4228d4da4492e4102db04865c
|
|
Change-Id: I431bfaa7722e4750d1fed828a40b71d23b3d5630
|
|
Remote copy & execution does not work, trying to understand why.
Change-Id: I1f8c94f7a691b1a6be857b5e3dd1f2851e0f2a37
|
|
Uses tcwg-tk1-01 to execute the generated program.
Change-Id: I7ac1dde8fab1f0f0f055b715046023715392d429
|
|
A fix was backported to the upstream gcc-5-branch and gcc-6-branch
branches. This means that since mid-March, both those branches
now produce a gcov-dump binary.
Change-Id: Id8bba80e942c5cc3462564186232276c06ba400d
|
|
We only started enabling gold for GCC 6 and later.
Change-Id: I85e133079e76bc166ebfe334f87d676569508da2
|
|
Change-Id: I451d7a1974e59ea9b334ec3f67c71a0a7a390576
|
|
Change-Id: Ib40250db1ccba839f36bc8953e13665f6cdb56d8
|
|
Change-Id: I0bb21cf05d79576a8ca25f53db3570d0d3d5428c
|
|
Change-Id: I3628b6d6e086378cf01e63680d95efa3fc8df256
|
|
Copy list.txt -> list-{5,6}.txt
Copy list-mingw.txt -> list-mingw-{5,6}.txt
Add gcov-dump to list.txt, list-mingw.txt for gcc7 (binutils 2.28)
Change-Id: Ica5e66a863ec108de59c86e6e4c7ac1c5feabdbd
|
|
Change-Id: I70fac8f96a7c4d78572229578ac3c441c01509fb
|
|
Checks there are no symlinks in mingw toolchains.
Change-Id: I4abc23e5a5e9f6d589e6b5ac2107ac1843d7890e
|
|
The bare metal toolchains do not provide gdbserver or gfortran.
Therefore they should be removed from the list of expected binaries
before comparison.
Change-Id: I09b8dc32a2e426c9692c64e703dd0d359cd5690d
|
|
GCC plugins are not supported under Windows.
Change-Id: I948f5756fab85ae2a03e9e5d2bd6f940a1455438
|
|
This acts as a regression test for
https://bugs.linaro.org/show_bug.cgi?id=2295
Change-Id: Iea3438d33d279ff1746833ea4f780476ef2d263a
|
|
GCC plugins should be supported by cross-compiler for bare-metal targets.
Change-Id: I35b4279b906a936c8ffaf226ff5edfa6c898926b
|
|
GCC plugins should be compiled by the host compiler.
common.mk: Define HOSTCC and HOSTCXX.
Change-Id: I1e5ecb3731a09f4310f7bec9d0c2089303ce3b1a
|
|
Change-Id: I93091d0cb23fa47c07cd8e3f8a146dadf6eb3433
|
|
Change-Id: Ied9a363672248735d90757831815103107c70fad
|
|
|
|
Build infrastructure, README, and basic tests for:
- hello.c and hello.cxx link
- toolchain properties (endianness, default Thumb code generation, default FPU, float-abi)
- tests for bugs 2034, 2123, 2764
|