Re: cvs2git and CVS keywords

Michael Haggerty wrote:
> Reto Glauser wrote:
>> I do have the following question: is it possible to ignore the CVS 
>> keywords in each file during import or is it necessary to create the 
>> branching history? IIRC git just looks at the content of a file and 
>> creates the SHA1 from it - so, if the content changes (which does of 
>> course in CVS in each different branch due to the keywords), we miss 
>> that a commit is the same as another. This results in only 1 return from:
>>
>> $ git branch --contains <SHA1>
>>
>> even though the same commit/file exists in 5 branches. Maybe it's 
>> possible with some grep/sed-magic to rewrite each file before commiting 
>> it to git?
> 
> Yes, it would be possible, with some minor changes to cvs2svn.  There is
> already code in cvs2svn that knows how to collapse or expand CVS
> keywords (see cvs2svn_lib/checkout_internal.py,
> InternalRevisionReader.get_content_stream()).  The handling of CVS
> keywords by cvs2git is not really thought out; see
> cvs2svn_lib/git_output_option.py,
> GitRevisionInlineWriter._modify_file(), so some work here would indeed
> be welcome.
> 
> Alternatively, one could use git-filter-branch with some random shell
> commands *after* the conversion.
> 

Turns out the default running operation for git is:

ctx.revision_recorder = SimpleFulltextRevisionRecorderAdapter( ... )

which means, no GitRevisionInlineWriter is called.

To *NOT* substitute the CVS keywords in this default mode, we can pass 
TRUE for suppress_keyword_substitution in the get_content_stream() call 
of record_text in SimpleFulltextRevisionRecorderAdapter which then 
passes '-kk' to CVS. This way, CVS won't hand over the files with 
expanded keywords and we're all set:

> $ git diff master -- cvs2svn_lib/fulltext_revision_recorder.py
> diff --git a/cvs2svn_lib/fulltext_revision_recorder.py b/cvs2svn_lib/fulltext_revision_recorder.py
> index ad057b7..94068d3 100644
> --- a/cvs2svn_lib/fulltext_revision_recorder.py
> +++ b/cvs2svn_lib/fulltext_revision_recorder.py
> @@ -114,8 +113,9 @@ class SimpleFulltextRevisionRecorderAdapter(FulltextRevisionRecorderAdapter):
>      # FIXME: We have to decide what to do about keyword substitution
>      # and eol_style here:
>      fulltext = self.revision_reader.get_content_stream(
> -        cvs_rev, suppress_keyword_substitution=False
> +        cvs_rev, suppress_keyword_substitution=True
>          ).read()
>      return self.fulltext_revision_recorder.record_fulltext(
>          cvs_rev, log, fulltext
>          )

------------------------------------------------------
http://cvs2svn.tigris.org/ds/viewMessage.do?dsForumId=1670&dsMessageId=2403240

To unsubscribe from this discussion, e-mail: [[email protected]].