Vacations.
CVS authors don't follow their own protocol. Particularly, they arbitrary
omit the space character from the sequence "ok \n" indicating the end of transmission.
Additional checking added to MyCVS in order to catch this bug.
Primitive conversation "Valid_responses - valid_requests" implemented in MyCVS.
Reading CVS info manual again. Whom this book is for? Most terms are heavily using before they are defined. Basic features, like "list files without getting their contents", are absent. Developer's crap, like "locks in the repository", is given in details. I have a temptation to write CVS help.
Handshake operation improved. Space character removed from most commands. Protocol documentation seems to be wrong. First checkout using MyCVS passed.
MyCVS command-line user interface designed. The program
will provide its own console prompt, similarly to widely used ftp and telnet.
Today's implementation processes only the command exit. MyCVS will
also support traditional CVS single-command mode for compatibility reasons.
Most MyCVS commands implemented: connect, disconnect,
help, pwd, and web.
MyCVS command ls (dir) implemented.
Windows API supporting locale settings is very inconvenient. For example, I want
to print the file size (merely the number of bytes) separating thousands with
the appropriate character. If I use "default locale" for US-English language
settings, Windows appends fractional decimal part (decimal dot and two zeroes)
to the integer number of bytes. In order to cut this tail, I have to use "custom
formatting": 90 lines of code just to print one number!
MyCVS commands cd, helpcvs, and cd implemented.
MyCVS own command opt must show the dialog box. I
don't want to use any framework library for that purpose. From other hand, I
want to encapsulate the dialog behavior in one C++ class. Therefore, I need to
associate my C++ class with the standard pair "window handle + window procedure".
The interesting solution, borrowed from ATL, is discussed in following two articles:
http://www.codeconduct.com/ATL/thunkwnd.asp
http://www.codeguru.com/atl/Thunk.html
The idea is simple. When the C++ class creates the window, it injects virtual
callback function into the window procedure (which is the inseparable part of
the window). The callback function passes input parameters from the window
procedure to the C++ class, handles them accordingly to C++ class behavior, and
returns the result to the window procedure caller.
The virtual callback mechanism applied to MyCVS Options dialog. The cost of using C++ class instead of dialog procedure is 100 lines of code.
Prototyping MyCVS Options dialog.
According to "Designed for Windows" recommendations, I need to provide uninstall functionality for MyCVS. Therefore, I should use Registry. I wrote the C++ class encapsulating work with Registry. Not surprisingly, it is similar to ATL CRegKey class, but doesn't depend from any ATL file.
I'm looking for prototypes of MyCVS window user interface design. SourceSafe uses different storing and checking technology. Its Linux analog Source OffSite is a nightmare ( I'm biased against it because I know it well). Cervisia has too long drop-down menus. WinCVS (or CvsGui; they cannot decide) has no design at all.
I did a sketch of the MyCVS user interface in window mode. By estimation, in order to release the product in February, I should postpone window interface to the next version.
MyCVS continues to grow. I finished low-level CVS protocol
primitives: Argument, Argumentx, Directory, Entry, Global_option,
Is_modified, Modified, Root, Unchanged, UseUnchanged.
MyCVS protocol implementation evolves.
I wrote following low-level CVS protocol
primitives: add, admin, annotate, ci, co, diff, editors, export, history,
import, init,
log, noop, rdiff, release, remove, rtag, status, tag, update, update_patches,
watch_on, watch_off, watch_add, watch_remove, watchers, valid_requests, version.
Too many commands.
I'm working on MyCVS at home, in rare free hours. I have to
take the maximum from every line of the program source. With that goal in mind,
I read "The C++ Programming Language" by B.Stroustrup. It's a treasure. It has
the same effect as ancient texts: the more you already know, the more you will
know after reading. Five years ago, the book seemed difficult. Two years ago, I
was excited. Today I admire.
I found dozens very interesting points (for example, the possibility to override
a private virtual function in derived classes).
I splitted one big MyCVS source file to twenty small class files. Most of classes are entirely contained in header files.
In order to use terms correctly, I re-read "Object-Oriented Analysis and Design" by G. Booch. In my opinion, there is more analysis than design. Therefore, the book is best suited for early development stages. Now MyCVS has a good theoretical background, and I'm ready to start conceptual modeling. Just like in most software projects, the design starts after writing code. I call it Over-Extreme Programming.
The article describing CVS protocol started. The best way to understand what is what is to write the manual. Most articles in popular computer magazines are written this way...