Oddity with cygport, debuginfo generation, and rxvt-unicode

Oddity with cygport, debuginfo generation, and rxvt-unicode

[Charles Wilson]

While rebuilding rxvt-unicode, I saw the following error during 'cygport
install':

Preparing debuginfo source files:
realpath: unknown option -- p
Try `realpath --help' for more information.
	80 files
Preparing executables for UAC:


A little debugging showed that realpath is being invoked during
src_postinstall here:

line 978:do
                 # make sure that paths with '../' don't leave srcdir
                 case $(realpath -m ${dbgsrc}) in
                         ${dbgsrcdir}/*) ;;
                         *)  continue ;;
                 esac

                 srcf=${dbgsrc#${dbgsrcdir}/}

I see that this line is hit many times, but here is the relevant sequence:

...
realpath -m <iom_perl.xs
realpath -m -pe
realpath: unknown option -- p
Try `realpath --help' for more information.
realpath -m s/IOM_MODULE/urxvt/g,s/IOM_CLASS/urxvt/g
realpath -m |
realpath -m /usr/src/debug/rxvt-unicode-X-9.07-2/src/../libev/ev++.h
...

Although only the second of these lines triggers the error message, each
of the first four is problematic.

This looks like a bug in gcc/binutils to me; debug src paths shouldn't
include pipe line shell commands, should they?

In the meantime, I think a fix like this for cygport would be appropriate:

line 978:do
+          # avoid invoking realpath with redirection metacharacters
+          case "${dbgsrc}" in
+          *[\<\|\>]* ) continue;
+          esac

           # make sure that paths with '../' don't leave $dbgsrcdir
-          case $(realpath -m ${dbgsrc}) in
+          case $(realpath -m -- ${dbgsrc}) in
                   ${dbgsrcdir}/*) ;;
                   *)  continue ;;
           esac

           srcf=${dbgsrc#${dbgsrcdir}/}

...although I'm not sure if this might cause a problem if a filename
actually contains a shell redirection metacharacter (and the originating
project is always careful to appropriate quote that filename).

--
Chuck

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Oddity with cygport, debuginfo generation, and rxvt-unicode

[Yaakov (Cygwin/X)]

On 2012-08-21 12:20, Charles Wilson wrote:
> While rebuilding rxvt-unicode, I saw the following error during 'cygport
> install':
>
> Preparing debuginfo source files:
> realpath: unknown option -- p
> Try `realpath --help' for more information.
> 	80 files
> Preparing executables for UAC:
>
>
> A little debugging showed that realpath is being invoked during
> src_postinstall here:
>
> line 978:do
>                   # make sure that paths with '../' don't leave srcdir
>                   case $(realpath -m ${dbgsrc}) in
>                           ${dbgsrcdir}/*) ;;
>                           *)  continue ;;
>                   esac
>
>                   srcf=${dbgsrc#${dbgsrcdir}/}

FWIW, readlink is now used in git instead of realpath, as the latter is 
not available in Fedora 16 or EL6, but the result should be the same.

> I see that this line is hit many times, but here is the relevant sequence:
>
> ...
> realpath -m <iom_perl.xs
> realpath -m -pe
> realpath: unknown option -- p
> Try `realpath --help' for more information.
> realpath -m s/IOM_MODULE/urxvt/g,s/IOM_CLASS/urxvt/g
> realpath -m |
> realpath -m /usr/src/debug/rxvt-unicode-X-9.07-2/src/../libev/ev++.h
> ...
>
> Although only the second of these lines triggers the error message, each
> of the first four is problematic.
>
> This looks like a bug in gcc/binutils to me; debug src paths shouldn't
> include pipe line shell commands, should they?

Not AFAIK.  That is very odd, but I am able to duplicate it with your 
current package.

> In the meantime, I think a fix like this for cygport would be appropriate:

Could you confirm that just this change fixes this:

-          case $(realpath -m ${dbgsrc}) in
+          case $(readlink -m -- "${dbgsrc}") in

> ...although I'm not sure if this might cause a problem if a filename
> actually contains a shell redirection metacharacter (and the originating
> project is always careful to appropriate quote that filename).

And when was the last time you saw that?


Yaakov


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Oddity with cygport, debuginfo generation, and rxvt-unicode

[Charles Wilson]

On 8/21/2012 4:48 PM, Yaakov (Cygwin/X) wrote:
> On 2012-08-21 12:20, Charles Wilson wrote:
>> In the meantime, I think a fix like this for cygport would be
>> appropriate:
>
> Could you confirm that just this change fixes this:
>
> - case $(realpath -m ${dbgsrc}) in
> + case $(readlink -m -- "${dbgsrc}") in

Yes, that change, by itself, fixes the problem.

>> ...although I'm not sure if this might cause a problem if a filename
>> actually contains a shell redirection metacharacter (and the originating
>> project is always careful to appropriate quote that filename).
>
> And when was the last time you saw that?

Oh, let me think...never. I'm still (re)training people that spaces are 
evil.  Don't even get me started on hashes and shell metas in filenames...

--
Chuck

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Oddity with cygport, debuginfo generation, and rxvt-unicode

[Earnie Boyd]

On Tue, Aug 21, 2012 at 11:40 PM, Charles Wilson wrote:
> Oh, let me think...never. I'm still (re)training people that spaces are
> evil.  Don't even get me started on hashes and shell metas in filenames...

Oh, just rename them to be asc(0x010); that's always good for a laugh. ;)

-- 
Earnie
-- https://sites.google.com/site/earnieboyd

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple