Does MAX_PATH issue still exists in Windows 10 - path

Can someone please tell us/me if the MAX_PATH issue still exists in (the technical preview of) Windows 10. And if it exists: How many characters can a path and an individual file name have?

The issue will be always present in Windows, to keep compatibility with old software. Use the NT-style name syntax "\\?\D:\very long path" to workaround this issue.
In Windows 10 (Version 1607 - Anniversary Update) and Windows Server 2016 you seem to have an option to ignore the MAX_PATH issue by overriding a group policy entry enable NTFS long paths under Computer Configuration -> Admin Templates -> System -> FileSystem:
The applications must have an entry longPathAware similar to DPIAware in the application manifest.
<assembly xmlns="urn:schemas-microsoft-com:asm.v1" manifestVersion="1.0" xmlns:asmv3="urn:schemas-microsoft-com:asm.v3" >
<asmv3:application>
<asmv3:windowsSettings xmlns="http://schemas.microsoft.com/SMI/2016/WindowsSettings">
<longPathAware>true</longPathAware>
</asmv3:windowsSettings>
</asmv3:application>
</assembly>

Here's some ansible code to enable long paths to avoid all that clicking in #magicandre1981 answer. This was tested on Windows Server 2016, it should work on Windows 10 too.
- name: Remove filesystem path length limitations
win_regedit:
path: HKLM:\SYSTEM\CurrentControlSet\Control\FileSystem
name: LongPathsEnabled
type: dword
data: 1
state: present

Yes it does still exist. Just ran into an issue now and the usual method of mapping a network drive to it to shorten the path didn't seem to let me open the files, but it would let me rename and move them.

Related

Continuous Integration with Blue Ocean, Github and Nuget causes path too long

NUnit.Extension.VSProjectLoader.3.7.0
I try to get a build chain to work with Jenkins Blue Ocean where the sources are in GitHub and additional dependencies are in nuget.
When I restore packages I get the error after the specific package NUnit.Extension.VSProjectLoader.3.7.0:
Errors in packages.config projects
The specified path, file name, or both are too long. The fully qualified file name must be less than 260 characters, and the directory name must be less than 248 characters.
On the agent machine the path is very short: C:\guinode\ on top of that additional length is added making the packages folder the following size:
MyGitProject is replacing my actual project name, the length is equal.
C:\guinode\workspace\MyGitProject_master-CFRRXMXQEUULVB4YKQOFGB65CQNC4U5VJKTARN2A6TSBK5PBATBA\packages
Checking the package on the agent machine shows that NUnit.Extension.VSProjectLoader.3.7.0 was loaded completely.
Checking a local installation and replacing the first path of the package I can find two files that are 260 characters or longer.
They belong to an internal project, so I have a chance of influencing that.
None of the directories are 248 characters or more.
So the immediate solution for me is to redeploy the internal reference package.
My question for future reference is if I can do something to the packages location or something to workspace\MyGitProject_master-CFRRXMXQEUULVB4YKQOFGB65CQNC4U5VJKTARN2A6TSBK5PBATBA so that I save some characters per default.
According to the microsoft documentation it can be possible to modify the 260 length rule.
If you prefix your file with '\\?\' eg: '\\?\C:\guinode\workspace...' then long path will be in use ( a little bit more than 32000 char). I hope settings JENKINS_HOME to this kind of path make all process use that (I'm not sure)
On recent Windows version (10.1607, 2016?) there is an option in the registry to enable long path. Set 1 to the following key: HKLM\SYSTEM\CurrentControlSet\Control\FileSystem LongPathsEnabled (Type: REG_DWORD) and restart the process.

result from /proc/self/exe is unfriendly in a clearcase view

If I execute a binary in a clearcase view, and look at /proc/self/exe for that on Linux, I see something like the following:
$ cd /proc/19220
$ ls -l exe
lrwxrwxrwx 1 peeterj pdxdb2 0 2012-11-30 13:04 exe -> /home/peeterj/views/peeterj_clang-7.vws/.s/00024/8000028250b8f1d1llvm-config
The clang llvm-config program, not unreasonably, uses this output to try to figure out the absolute fully qualified path that it is located in (I assume in case argv[0] isn't fully qualified).
Is there a way to find the location within the view that this corresponds to. For example, in this case, the llvm-config exe is actually in:
/vbs/bldsupp/linuxamd64/clang/debug/bin
(I'm wondering if it's feasible to modify clang's GetExecutablePath() function to figure this out.)
No trivial solution here (for an old version of ClearCase though):
The technote "PK27447: WITHIN A CLEARCASE DYNAMIC VIEW, THE READLINK() CALL ON LINUX RETURNS THE WRONG PATH FOR THE EXECUTABLE'S /PROC/SELF/EXE VALUE" suggests:
Local fix
Use getcwd(), get_current_dir_name(), getwd() in applications slated for a VOB/View context
Create an interposer library to intercept the readlink() call, and modify to use any of the above calls to return the proper data
The cause:
/proc/self/exe returns the improper path while getcwd succeeds.
Unfortunately, for /proc/self/exe to return the proper value [from within a VOB/View context] would require a change within the Linux kernel to allow MVFS to "override" the present setting.
IBM LTC has been working on having the Linux community adopt this change so that we can then incorporate the new features within MVFS.
Related: Bug Sun 6189256.

Deleting .inf and .pnf Files

I manually install my driver using an .inf file. Until now, I deleted the oem.inf and .pnf files from the inf folder to uninstall. Whenever I install a new driver I delete/uninstall the old inf and pnf files.
In my old uninstalls (by deleting .inf and .pnf files), I didn't modify or delete any registry settings.
In this case do I want to change or remove any settings from the registry (for example: devnode)?
You should use the SetupUninstallOEMInf function to uninstall the .INF (and subsequently .PNF) files. This will take care of the details. pnputil (on Vista and higher) should do the equivalent thing from the command line. However, this function will not delete drivers that are currently installed (e.g. associated with a devnode).
Why are you uninstalling the old driver first? The user might already installed your driver for at least one devnode. Why not use a Microsoft-sanctioned solution such as DpInst? It will do the work required to update the driver.
Passing SUOI_FORCEDELETE to SetupUninstallOEMInf wouldn't be a good idea, cause you'd end up with lingering .INF references in your devnodes (in the registry).
At work I wrote a utility I called DriverUninstaller that deletes the devnodes and then deleted the INFs. I only use this utility for uninstallations. Upgrades are handled by DpInst, as they should be. The flow is roughly:
Enumerate them with SetupAPI (e.g. by device class if your device class is unique)
For each devnode, call SetupDiCallClassInstaller(DIF_REMOVE, ...)
Call SetupDiBuildDriverInfoList to find all .INF files for my device
For each INF, call SetupUninstallOEMInf
If there'll be interest in this utility, I might be able to persuade my employer to open-source it :-)
As the other answer points out the API to remove drivers on Windows is the SetupUninstallOEMInf method but I figured I'd add a few important notes here:
the INF path parameter must be the file name only (must not include the full path!)
this API requires Administrator privileges (this is not much of a surprise)
on 64-bit systems, the method only works when executed in the context of a 64-bit process (ie. WOW64 doesn't work)
The SetupDiGetDriverInfoDetail API can be used to query information about the relevant INF file(s) to remove for a particular device. And the SetupDiEnumDriverInfo/SetupDiBuildDriverInfoList APIs can be used to enumerate all drivers for a particular device.
// given infFilePath - full path to inf as returned by a query using SetupDiGetDriverInfoDetail
TCHAR* infFileName = GetFileNamePart(infFilePath);
if(SetupUninstallOEMInf(pInf, SUOI_FORCEDELETE, NULL))
{
// success
}else
{
DWORD errCode = GetLastError();
if(errCode == 0x02)
{
// means that the driver INF file was not found
// most likely it was already uninstalled
}else if(errCode == 0x05)
{
// must run as administrator
}else
{
// some other error code.. handle appropriately
}
}
I've open-sourced the code for a tool I wrote to perform driver uninstalls for USB and Media/Image devices. Details here: http://mdinescu.com/software-development/30-driver-hunter-programatically-uninstall-drivers-in-windows

How to get Excel version and macro security level

Microsoft has recently broken our longtime (and officially recommended by them) code to read the version of Excel and its current omacro security level.
What used to work:
// Get the program associated with workbooks, e.g. "C:\Program Files\...\Excel.exe"
SHELLAPI.FindExecutable( 'OurWorkbook.xls', ...)
// Get the version of the .exe (from it's Properties...)
WINDOWS.GetFileVersionInfo()
// Use the version number to access the registry to determine the security level
// '...\software\microsoft\Office\' + VersionNumber + '.0\Excel\Security'
(I was always amused that the security level was for years in an insecure registry entry...)
In Office 2010, .xls files are now associated with "“Microsoft Application Virtualization DDE Launcher," or sftdde.exe. The version number of this exe is obviously not the version of Excel.
My question:
Other than actually launching Excel and querying it for version and security level (using OLE CreateOLEObject('Excel.Application')), is there a cleaner, faster, or more reliable way to do this that would work with all versions starting with Excel 2003?
Use
function GetExcelPath: string;
begin
result := '';
with TRegistry.Create do
try
RootKey := HKEY_LOCAL_MACHINE;
if OpenKey('SOFTWARE\Microsoft\Windows\CurrentVersion\App Paths\excel.exe', false) then
result := ReadString('Path') + 'excel.exe';
finally
Free;
end;
end;
to get the full file name of the excel.exe file. Then use GetFileVersionInfo as usual.
As far as I know, this approach will always work.
using OLE CreateOLEObject('Excel.Application'))
you can get installed Excel versions by using the same registry place, that this function uses.
Basically you have to clone a large part of that function registry code.
You can spy on that function call by tools like Microsoft Process Monitor too see exactly how does Windows look for installed Excel - and then to do it exactly the same way.
You have to open registry at HKEY_CLASSES_ROOT\ and enumerate all the branches, whose name starts with "Excel.Application."
For example at this my workstation I only have Excel 2013 installed, and that corresponds to HKEY_CLASSES_ROOT\Excel.Application.15
But on my another workstation I have Excel 2003 and Excel 2010 installed, testing different XLSX implementations in those two, so I have two registry keys.
HKEY_CLASSES_ROOT\Excel.Application.12
HKEY_CLASSES_ROOT\Excel.Application.14
So, you have to enumerate all those branches with that name, dot, and number.
Note: the key HKEY_CLASSES_ROOT\Excel.Application\CurVer would have name of "default" Excel, but what "default" means is ambiguous when several Excels are installed. You may take that default value, if you do not care, or you may decide upon your own idea what to choose, like if you want the maximum Excel version or minimum or something.
Then when for every specific excel branch you should read the default key of its CLSID sub-branch.
Like HKEY_CLASSES_ROOT\Excel.Application.15\CLSID has nil-named key equal to
{00024500-0000-0000-C000-000000000046} - fetch that index to string variable.
Then do a second search - go into a branch named like HKEY_CLASSES_ROOT\CLSID\{00024500-0000-0000-C000-000000000046}\LocalServer ( use the fetched index )
If that branch exists - fetch the nil-named "default key" value to get something like C:\PROGRA~1\MICROS~1\Office15\EXCEL.EXE /automation
The last result is the command line. It starts with a filename (non-quoted in this example, but may be in-quotes) and is followed by optional command line.
You do not need command line, so you have to extract initial commanlind, quoted or not.
Then you have to check if such an exe file exists. If it does - you may launch it, if not - check the registry for other Excel versions.

SaveDialog.Execute not doing anything in Windows 7

Delphi 2007 on windows 7 just does nothing on the saveDialog.Execute call. I have seen another person mention this a few weeks back but it was with Borland c++.
See the thread "TOpenDialog.Execute not working " on embarcadero newsgroups.
Problem there was resolved by deleting the executable name from
"HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options"
I got the same problem (savedialog not working) in windows XP.
After lots of unsuccessful attempts according to the voluminous exchanges in the embarcadero group you mention (https://forums.embarcadero.com/thread.jspa?messageID=196950&tstart=0#196950).
I found what the reason was : the initial file dir and filename of the Savedialog12 were bad, contradicting each other, the filename containing the fullpath of the last file I had opened (I had thought it was smart to prepare the saving of the file I had opened last; unfortunately what I had put in the initialdir was equal to what I had put in the filename !)
The problem was already solved by clearing both fields of the save dialog.
Further, my initial goal to prepare the saving was reached by putting valid values in the involved fields :
SaveDialog1.FileName:=ExtractFileName(Opendialog1.Filename);
SaveDialog1.InitialFileDir:=ExtractFilePath(Opendialog1.Filename);

Resources