Ros-dev

ros-dev@reactos.org

5788 discussions

by Colin Finck

Hi all! A picture says more than a thousand words: http://fs5.directupload.net/images/170613/u4d8a59a.png As promised in the last meeting, I have brought to life a machine for testing GPU drivers with real GPUs under ReactOS. It combines a QEMU virtual machine with PCIe GPU Passthrough together with a KVM-over-IP card for feeding the real video signal back to a window. What you see in the screenshot is a real NVIDIA GeForce 9300 GPU attached to a ReactOS VM. We can now test the NVIDIA driver installation while enjoying the easy debuggability of a VM. My first report is already out: https://jira.reactos.org/browse/CORE-13423 Every developer interested in working on this can drop me a line and I will send you credentials for the server. Soon I will also add an AMD Graphics Card along with another KVM-over-IP card to let us debug the other big driver package. Thanks go out to Mr. Kromdijk for the generous server donation, Christoph for the KVM-over-IP card, and Hervé for the QEMU hints! Let's all get some flashy 3D acceleration into ReactOS!! :) Cheers, Colin

7 years, 6 months

Re: [ros-dev] [ros-diffs] [hbelusca] 75008: [USETUP]: Introduce SetupDeleteFile() and SetupMoveFile() (in addition to the already-existing SetupCopyFile()) in order to implement moving / renaming existing files. Will be use...

by Thomas Faber

On 2017-06-12 01:58, hbelusca(a)svn.reactos.org wrote: > +NTSTATUS > +SetupMoveFile( > + IN PCWSTR ExistingFileName, > + IN PCWSTR NewFileName, > + IN ULONG Flags) > +{ > + NTSTATUS Status; > + IO_STATUS_BLOCK IoStatusBlock; > + OBJECT_ATTRIBUTES ObjectAttributes; > + PFILE_RENAME_INFORMATION RenameInfo; > + UNICODE_STRING NewPathU, ExistingPathU; > + HANDLE SourceHandle = NULL; This variable fulfills the requirements for _SEH2_VOLATILE > + BOOLEAN ReplaceIfExists; > + > + RtlInitUnicodeString(&ExistingPathU, ExistingFileName); > + RtlInitUnicodeString(&NewPathU, NewFileName); > + > + _SEH2_TRY > + { Which part of this code needs SEH? That looks like it's purely hiding bugs.

7 years, 6 months

Re: [ros-dev] [ros-diffs] [tthompson] 74970: [NTFS] - Add the most basic support for file creation. Expand diagnostic output, especially in NtfsDumpIndexRootAttribute(). Replace an ExFreePool() with ExFreePoolWithTag(). Add...

by Thomas Faber

On 2017-06-09 05:14, tthompson(a)svn.reactos.org wrote: > + RtlZeroMemory(&FilenameNoPath, sizeof(UNICODE_STRING)); > + FilenameNoPath.Buffer = Buffer; > + FilenameNoPath.MaximumLength = MAX_PATH; > + RtlInitEmptyUnicodeString does this for you On 2017-06-09 05:14, tthompson(a)svn.reactos.org wrote: > + FileNameAttribute->NameLength = FilenameNoPath.Length / 2; sizeof(WCHAR)?

7 years, 6 months

Re: [ros-dev] [ros-diffs] [phater] 74750: [MSWSOCK] Call explicitly DnsQuery_W. Copy host aliases if available as answer from the query

by Thomas Faber

On 2017-06-02 19:52, phater(a)svn.reactos.org wrote: > + aDstNextStr = (char*)((DWORD)aDstNextStr + (DWORD)bItmLen); DWORD cannot always fit a pointer. You want DWORD_PTR for this kind of math (or simply use char *).

7 years, 7 months

Re: [ros-dev] [ros-diffs] [hbelusca] 74742: [USETUP]: Creating a registry key with zero access is just plainly invalid. Fix it. It was not caught before because, we never ever booted a 1st-stage setup with a valid mounted e...

by Thomas Faber

On 2017-06-02 02:44, hbelusca(a)svn.reactos.org wrote: > @@ -341,7 +341,7 @@ > } > > InitializeObjectAttributes(&ObjectAttributes, &ServicesU, OBJ_CASE_INSENSITIVE, NULL, NULL); > - Status = NtCreateKey(&hServices, 0, &ObjectAttributes, 0, NULL, 0, NULL); > + Status = NtCreateKey(&hServices, KEY_ALL_ACCESS, &ObjectAttributes, 0, NULL, 0, NULL); > if (!NT_SUCCESS(Status)) > { > DPRINT1("NtCreateKey('%wZ') failed with status 0x%08lx\n", &ServicesU, Status); > > It doesn't really need full access, does it? Could have gone for something closer to the original, like READ_CONTROL.

7 years, 7 months

Re: [ros-dev] [ros-diffs] [hbelusca] 74741: [MKHIVE][CMAKE]: Make mkhive a bit more flexible, so that it can generate only specific hives on-demand (and not all of them always at once). I need this for building a single boo...

by Thomas Faber

I understand not making the change to how 1st stage deals with the registry in trunk. Why not the tool change though? Can't you have the new mkhive do the exact same thing it does now, and do it in trunk? The smaller your branch merge commit the easier it will be to avoid/pinpoint/fix regressions. On 2017-06-02 02:34, hbelusca(a)svn.reactos.org wrote: > Author: hbelusca > Date: Fri Jun 2 00:34:10 2017 > New Revision: 74741 > > URL: http://svn.reactos.org/svn/reactos?rev=74741&view=rev > Log: > [MKHIVE][CMAKE]: Make mkhive a bit more flexible, so that it can generate only specific hives on-demand (and not all of them always at once). I need this for building a single bootcd registry hive. > I commit these changes in my branch because it's too much über-advanced code for our trunk, yet... (and it will uncover deep "setup" hacks in NTOS' iomgr & pnpmgr as soon as I'll enable 1-st stage setup to have a proper registry present as done on windows). > CORE-13347 #comment Committed in r74741 but just in the setup_improvements branch for the moment.

7 years, 7 months

Re: [ros-dev] [ros-diffs] [hbelusca] 74733: [NTOS]: Minor code formatting + don't leak a handle in case of failure, in CmpInitializeHardwareConfiguration().

by Thomas Faber

On 2017-06-01 20:27, hbelusca(a)svn.reactos.org wrote: > /* EISA */ > case EisaAdapter: > - > + { > /* Fixup information */ > Interface = Eisa; > Bus = CmpTypeCount[EisaAdapter]++; > break; > + } This is a completely unnecessary change to somebody else's code that has no basis in our documented coding style. Uh... don't do that?

7 years, 7 months

Re: [ros-dev] [ros-diffs] [hbelusca] 74739: [MKHIVE]: Formatting changes only + sync back the names of the reg-inf functions with the ones where they are coming from (aka. Wine's setupapi/install.c).

by Thomas Faber

On 2017-06-02 02:00, hbelusca(a)svn.reactos.org wrote: > [MKHIVE]: Formatting changes only + sync back the names of the reg-inf functions with the ones where they are coming from (aka. Wine's setupapi/install.c). Those two goals clearly conflict with either other. Either keep Wine's function names AND formatting to keep things comparable/sync-able, or make the code follow our style.

7 years, 7 months

Re: [ros-dev] [ros-diffs] [hbelusca] 74735: [NTOS]: Better clean up behind ourselves on failure paths (aka. closing handles). These functions are clearly not written with correct cleanup-on-failure concept in mind...

by Thomas Faber

On 2017-06-01 20:37, hbelusca(a)svn.reactos.org wrote: > + if (!NT_SUCCESS(Status)) > + { > + /* We failed, close all the opened handles and return */ > + // NtClose(KeyHandle); Please don't add commented-out code. What is that supposed to even mean? "I was too lazy to test this"? "All the other code depends on this leak"? "I think it would look prettier here"?

7 years, 7 months

Status Meeting - rescheduled

by Aleksey Bragin

Hello, I propose to reschedule our monthly meeting to Thursday next week (1st of June), as tomorrow is a holiday in some countries, and also I have tight meetings schedule this week. Regards, Aleksey Bragin

7 years, 7 months

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Ros-dev