Issue patching sources when trying target darwin after previous patching successful.

[martell]

Lets say we build for i686-darwin target after extracting and patching sources and we fail because LLVM need's python to be installed. ie. > 2.5

When we add the missing package via pacman and try to continue with ct-ng build we get the following error.

[INFO ] Extract kernel headers and libraries (minimal): done in 103.27s (at 04:07)
[04:07] / mv: cannot stat '/home/Martell/cttest/.build/src/ld64-127.2/*': No such file or directory
[ERROR]
[ERROR] >>
[ERROR] >> Build failed in step 'Extracting and patching toolchain components'
[ERROR] >> called in step '(top-level)'

Maybe this has to do with the fact that we are time stamping to 1999-12-31

It's quite frustrating to have to delete the src directory everytime only to re extract and patch the sources again

[mingwandroid]

I agree it is annoying and we must fix it. The problem is because of how I
make a composite project for cctools and ld64 so as to cut down on the
overall patching burden (they share autoconf scripts). If we can't make it
work nicely in this scenario then I will have to keep them separate.

FWIW I spoke with Yann Morin about some closely related matters (Apple SDK
header copying into cctools via a patch) and he would rather we copied them
in at build time so that would also be a location in which to fix this
problem.
On Nov 24, 2013 10:00 PM, "martell" [email protected] wrote:

Lets saw we build for i686 darwin target after extracting and patching
sources and we fail because clang need python to be installed. ie. > 2.5

When we add the missing package via pacman and try to continue with ct-ng
build we get the following error.

[INFO ] Extract kernel headers and libraries (minimal): done in 103.27s
(at 04:07)
[04:07] / mv: cannot stat '/home/Martell/cttest/.build/src/ld64-127.2/*':
No such file or directory
[ERROR]
[ERROR] >>
[ERROR] >> Build failed in step 'Extracting and patching toolchain
components'
[ERROR] >> called in step '(top-level)'

Maybe this has to do with the fact that we are time stamping to 1999-12-31

It's quite frustrating to have to delete the src directory everytime only
to re extract and patch the sources again

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/4
.

[martell]

Ahh I see that explains it.

I don't think I would be best suited to fix this issue as I am only just starting to learn my way around the code base. I'd like to do a few more smaller fixes and adjustments before I start messing with the build scripts that much.

I have submitted a pull request for the other issue I was talking about earlier I believe that was the best introduction to the code base after the very simple patch update I done the last day
Slowly but surely I hope I become of some use to you guys.

I'll catch you on IRC at some point on Tuesday if your around to discuss further.

[mingwandroid]

That's fine! I'm happy to fix it but I want to get multilibs fixed first.

Thanks for the 3.4 patch. One minor thing, I wonder if passing an empty
string "" as branch to CT_GitGet could be used to indicate to use the old
hour clone command.. Or is -b "" equivalent anyway?

I am traveling back to UK today but won't be home until around 8pm do I
will log onto irc (#crosstool-ng) at that time.
On Nov 24, 2013 11:03 PM, "martell" [email protected] wrote:

Ahh I see that explains it.

I don't think I would be best suited to fix this issue as I am only just
starting to learn my way around the code base. I'd like to do a few more
smaller fixes and adjustments before I start messing with the internals
that much.

I have submitted a pull request for the other issue I was talking about
earlier I believe that was the best introduction to the code base after the
very simple patch update I done the last day
Slowly but surely I hope I become of some use to you guys.

I'll catch you on IRC at some point tomorrow if your around to discuss
further.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/4#issuecomment-29169076
.

[martell]

Passing in an empty string will result in an error.
We could check if it is empty and then do the old command.
The advantage of forcing a user to designate a branch does however decrease the clone and checkout time dramitically

I can add support for a blank argument if you want?

[mingwandroid]

I think support for the old behavior if branch is blank would make it
easier at review time; I'm not a git expert, is it possible to ask the
server what the name of the default clone branch is before cloning? If so
we could use this to initialize "branch" before running your new version
instead as I agree that the clone time is quite a pain otherwise. This
function wasn't written by us but I think we are the only ones who call it!

I also don't like that it clones to .build/tarballs and would rather it
cloned to $HOME/src in-line with what the other download functions do but
Yann Morin doesn't want this changed unfortunately.
On Nov 25, 2013 12:44 PM, "martell" [email protected] wrote:

Passing in an empty string will result in an error.
We could check if it is empty and then do the old command?
The advantage of forcing a user to designate a branch does however
decrease the clone and checkout time dramitically

I can add support for a blank argument if you want?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/4#issuecomment-29198622
.