gitignore pattern matching issue with Windows

gitignore pattern matching issue with Windows

[TDD Guru]

  Git does not seem to properly handle the pattern [AB] but it does handle [ab].

Here is the demonstration of the bug:

C:\gitbug>git --version
git version 2.50.1.windows.1

C:\gitbug>echo [ab] > .gitignore

C:\gitbug>type .gitignore
[ab]

C:\gitbug>git check-ignore -v a
.gitignore:1:[ab]       a

C:\gitbug>git check-ignore -v b
.gitignore:1:[ab]       b

C:\gitbug>git check-ignore -v A
.gitignore:1:[ab]       A

C:\gitbug>git check-ignore -v B
.gitignore:1:[ab]       B

C:\gitbug>echo [AB] > .gitignore

C:\gitbug>type .gitignore
[AB]

C:\gitbug>git check-ignore -v a

C:\gitbug>git check-ignore -v b

C:\gitbug>git check-ignore -v A

C:\gitbug>git check-ignore -v B

C:\gitbug>git config core.ignorecase
true

C:\gitbug>ver

Microsoft Windows [Version 10.0.19045.6093]

C:\gitbug>

Rex McCarthy

Re: gitignore pattern matching issue with Windows

[Junio C Hamano]

TDD Guru <tddguru@gmail.com> writes:

> C:\gitbug>git check-ignore -v a
>
> C:\gitbug>git check-ignore -v b
>
> C:\gitbug>git check-ignore -v A
>
> C:\gitbug>git check-ignore -v B

I do not think this is not limited to Windows.  The exclusion
patterns use the wildmatch machinery, and its ignore-case behaviour
seems iffy.  The code expects that pattern characters to be
lowercase in character classes in order to match their uppercase
counterpart, which sounds like a bug.

    $ cat >.gitignore <<\EOF
    [ab]
    [CD]
    EOF
    $ git -c core.ignorecase check-ignore -v A
    $ git -c core.ignorecase check-ignore -v C

Here is a totally untested patch to downcase the pattern characters
to match against the text characters that are already downcased.

 wildmatch.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git c/wildmatch.c w/wildmatch.c
index 69a2ae7000..1a2d86b691 100644
--- c/wildmatch.c
+++ w/wildmatch.c
@@ -189,10 +189,14 @@ static int dowild(const uchar *p, const uchar *text, unsigned int flags)
 			do {
 				if (!p_ch)
 					return WM_ABORT_ALL;
+				if ((flags & WM_CASEFOLD) && ISUPPER(p_ch))
+					p_ch = tolower(p_ch);
 				if (p_ch == '\\') {
 					p_ch = *++p;
 					if (!p_ch)
 						return WM_ABORT_ALL;
+					if ((flags & WM_CASEFOLD) && ISUPPER(p_ch))
+						p_ch = tolower(p_ch);
 					if (t_ch == p_ch)
 						matched = 1;
 				} else if (p_ch == '-' && prev_ch && p[1] && p[1] != ']') {
@@ -201,6 +205,8 @@ static int dowild(const uchar *p, const uchar *text, unsigned int flags)
 						p_ch = *++p;
 						if (!p_ch)
 							return WM_ABORT_ALL;
+						if ((flags & WM_CASEFOLD) && ISUPPER(p_ch))
+							p_ch = tolower(p_ch);
 					}
 					if (t_ch <= p_ch && t_ch >= prev_ch)
 						matched = 1;
@@ -225,6 +231,11 @@ static int dowild(const uchar *p, const uchar *text, unsigned int flags)
 							matched = 1;
 						goto next;
 					}
+
+					/*
+					 * [[:alnum:]] etc. must be spelled in
+					 * lowercase even under icase matching rule.
+					 */
 					if (CC_EQ(s,i, "alnum")) {
 						if (ISALNUM(t_ch))
 							matched = 1;