According to: https://docs.microsoft.com/en-us/windows/desktop/debug/pe-format#optional-he...
SizeOfHeaders: The combined size of an MS-DOS stub, PE header, and section headers rounded up to a multiple of FileAlignment.
This change also fixes two kernel loader tests currently failing under 64bit.
Signed-off-by: Brendan McGrath brendan@redmandi.com --- This patch set fixes three of the four failing 64bit kernel32 loader tests.
The remaining test that fails is due to the clr header being version 2.4 and wine only offering 64bit support for clr header version 2.5 and above. I couldn't find any documentation to state why that is so I've left the test as failing.
dlls/kernel32/tests/loader.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/dlls/kernel32/tests/loader.c b/dlls/kernel32/tests/loader.c index 01c7cfa8a911..aaaa39f601d4 100644 --- a/dlls/kernel32/tests/loader.c +++ b/dlls/kernel32/tests/loader.c @@ -1097,7 +1097,7 @@ static void test_Loader(void) nt64.OptionalHeader.MajorOperatingSystemVersion = 4; nt64.OptionalHeader.MajorImageVersion = 1; nt64.OptionalHeader.MajorSubsystemVersion = 4; - nt64.OptionalHeader.SizeOfHeaders = sizeof(dos_header) + sizeof(nt64) + sizeof(IMAGE_SECTION_HEADER); + nt64.OptionalHeader.SizeOfHeaders = ALIGN_SIZE(sizeof(dos_header) + sizeof(nt64) + sizeof(IMAGE_SECTION_HEADER), nt64.OptionalHeader.FileAlignment); nt64.OptionalHeader.SizeOfImage = nt64.OptionalHeader.SizeOfHeaders + 0x1000; nt64.OptionalHeader.Subsystem = IMAGE_SUBSYSTEM_WINDOWS_CUI; nt64.OptionalHeader.SizeOfStackReserve = 0x321000; @@ -1212,7 +1212,7 @@ static void test_Loader(void) nt32.OptionalHeader.MajorOperatingSystemVersion = 4; nt32.OptionalHeader.MajorImageVersion = 1; nt32.OptionalHeader.MajorSubsystemVersion = 4; - nt32.OptionalHeader.SizeOfHeaders = sizeof(dos_header) + sizeof(nt32) + sizeof(IMAGE_SECTION_HEADER); + nt32.OptionalHeader.SizeOfHeaders = ALIGN_SIZE(sizeof(dos_header) + sizeof(nt32) + sizeof(IMAGE_SECTION_HEADER), nt32.OptionalHeader.FileAlignment); nt32.OptionalHeader.SizeOfImage = nt32.OptionalHeader.SizeOfHeaders + 0x1000; nt32.OptionalHeader.Subsystem = IMAGE_SUBSYSTEM_WINDOWS_CUI; nt32.OptionalHeader.SizeOfStackReserve = 0x321000;
If the 32BITREQUIRED flag is present in the COM image header, then the image should not be loaded under a 64bit process.
This patch makes changes to an existing (failing) test to reflect that (it now passes).
Signed-off-by: Brendan McGrath brendan@redmandi.com --- dlls/kernel32/tests/loader.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/dlls/kernel32/tests/loader.c b/dlls/kernel32/tests/loader.c index aaaa39f601d4..44e1ef1cac6f 100644 --- a/dlls/kernel32/tests/loader.c +++ b/dlls/kernel32/tests/loader.c @@ -522,13 +522,15 @@ static NTSTATUS map_image_section( const IMAGE_NT_HEADERS *nt_header, const IMAG mod = LoadLibraryExA( dll_name, 0, DONT_RESOLVE_DLL_REFERENCES ); if (!has_code && nt_header->OptionalHeader.Magic == IMAGE_NT_OPTIONAL_HDR32_MAGIC) { - BOOL il_only = FALSE; + BOOL il_only = FALSE, il_32bit_only = FALSE; if (((const IMAGE_NT_HEADERS32 *)nt_header)->OptionalHeader.DataDirectory[IMAGE_DIRECTORY_ENTRY_COM_DESCRIPTOR].VirtualAddress) { const IMAGE_COR20_HEADER *cor_header = section_data; il_only = (cor_header->Flags & COMIMAGE_FLAGS_ILONLY) != 0; + il_32bit_only = (cor_header->Flags & COMIMAGE_FLAGS_32BITREQUIRED) != 0; } - ok( mod != NULL || broken(il_only), /* <= win7 */ + ok( mod != NULL || broken(il_only) /* <= win7 */ + || (is_win64 && il_32bit_only && mod == NULL), "%u: loading failed err %u\n", line, GetLastError() ); } else
Hi,
While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check?
Full results can be found at: https://testbot.winehq.org/JobDetails.pl?Key=45168
Your paranoid android.
=== debian9 (64 bit WoW report) ===
kernel32: loader.c:532: Test failed: 1275: loading failed err 193
Brendan McGrath brendan@redmandi.com writes:
According to: https://docs.microsoft.com/en-us/windows/desktop/debug/pe-format#optional-he...
SizeOfHeaders: The combined size of an MS-DOS stub, PE header, and section headers rounded up to a multiple of FileAlignment.
This change also fixes two kernel loader tests currently failing under 64bit.
Signed-off-by: Brendan McGrath brendan@redmandi.com
This patch set fixes three of the four failing 64bit kernel32 loader tests.
The remaining test that fails is due to the clr header being version 2.4 and wine only offering 64bit support for clr header version 2.5 and above. I couldn't find any documentation to state why that is so I've left the test as failing.
These tests succeed on Windows. You should be fixing the Wine implementation to pass the tests, not the other way around.
Apologies - I didn't realise this test passed on Windows.
I do have a patch that will fix this in the wine implementation; but I felt it was a bit unwieldy for such a niche scenario (where the PE32 header couldn't be remapped on to a PE32+ header). But I'm happy to submit it if that's the preferred approach.
On 29/11/18 8:33 pm, Alexandre Julliard wrote:
Brendan McGrath brendan@redmandi.com writes:
According to: https://docs.microsoft.com/en-us/windows/desktop/debug/pe-format#optional-he...
SizeOfHeaders: The combined size of an MS-DOS stub, PE header, and section headers rounded up to a multiple of FileAlignment.
This change also fixes two kernel loader tests currently failing under 64bit.
Signed-off-by: Brendan McGrath brendan@redmandi.com
This patch set fixes three of the four failing 64bit kernel32 loader tests.
The remaining test that fails is due to the clr header being version 2.4 and wine only offering 64bit support for clr header version 2.5 and above. I couldn't find any documentation to state why that is so I've left the test as failing.
These tests succeed on Windows. You should be fixing the Wine implementation to pass the tests, not the other way around.
Brendan McGrath brendan@redmandi.com writes:
Apologies - I didn't realise this test passed on Windows.
I do have a patch that will fix this in the wine implementation; but I felt it was a bit unwieldy for such a niche scenario (where the PE32 header couldn't be remapped on to a PE32+ header). But I'm happy to submit it if that's the preferred approach.
Yes that's what you have to do, changing the tests to accomodate Wine bugs is not allowed.
-
Alexandre Julliard -
Brendan McGrath -
Marvin