BrowserUk has asked for the wisdom of the Perl Monks concerning the following question:
This is effectively a follow-up to the issue I raised in Re^4: Missing bufferoverflowU.lib.
Later versions of MSVC compiler suites require an additional step in the build process to attach a 'manifest' resource to executable binaries (.exes & .dlls). The manifest files themselves are produced by the linker automatically. All that is required is to attach those files to the binaries (also produced by the linker) using mt.exe
(Quite why the linker can't do this for us automatically is moot question!).
Without this step being invoked, any attempt to load the dlls results in runtime error:
R6034: An application has made an attempt to load the C runtime librar +y incorrectly.
The problem is, that existing EU::MM (and similar) produced makefiles have a 'manifest' target and the command to invoke mt.exe to perform the attachment, but that target does not have any dependancies and so is never invoked.
Whilst I can hack my way past the problem for individual makefiles--by modifying the makefile before using it--that is not a good solution when attempting to install complex packages with many dependancies. Using CPAN for example.
Presumably, as the 'manifest' target is being produced by (most?) makefile.pls, the correct place to fix this is somewhere inside the EU::MM stuff. Does anyone have any cluebats as to how to go about this?
|
|---|
| Replies are listed 'Best First'. | |
|---|---|
|
Re: A permanent fix for the Windows 'manifest' issue?
by Corion (Patriarch) on Mar 23, 2009 at 15:13 UTC | |
by BrowserUk (Patriarch) on Mar 23, 2009 at 16:05 UTC | |
by ikegami (Patriarch) on Mar 23, 2009 at 16:58 UTC | |
by BrowserUk (Patriarch) on Mar 23, 2009 at 17:10 UTC | |
by admiral_grinder (Pilgrim) on Mar 24, 2009 at 19:23 UTC |