summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorEdward Thomson <ethomson@edwardthomson.com>2018-12-03 13:35:30 -0800
committerEdward Thomson <ethomson@edwardthomson.com>2018-12-03 17:02:10 -0800
commitef8f8ec604035142916271178b009f57faf1877a (patch)
treeb5836ab7f84946055777ca5c9c1aa0c207215410
parent59b054cb7efbfa97153f3d4dc0ab22c9ded31a22 (diff)
downloadlibgit2-ef8f8ec604035142916271178b009f57faf1877a.tar.gz
crlf: update to match git's logic
Examine the recent CRLF changes to git by Torsten Bögershausen and include similar changes to update our CRLF logic to match. Note: Torsten Bögershausen has previously agreed to allow his changes to be included in libgit2.
-rw-r--r--src/buf_text.c3
-rw-r--r--src/crlf.c403
-rw-r--r--src/filter.h10
3 files changed, 215 insertions, 201 deletions
diff --git a/src/buf_text.c b/src/buf_text.c
index 306980b5c..fa7f16b81 100644
--- a/src/buf_text.c
+++ b/src/buf_text.c
@@ -310,6 +310,7 @@ bool git_buf_text_gather_stats(
}
}
- return (stats->nul > 0 ||
+ /* Treat files with a bare CR as binary */
+ return (stats->cr != stats->crlf || stats->nul > 0 ||
((stats->printable >> 7) < stats->nonprintable));
}
diff --git a/src/crlf.c b/src/crlf.c
index 96cd2d807..ab2419c4c 100644
--- a/src/crlf.c
+++ b/src/crlf.c
@@ -18,68 +18,58 @@
#include "buf_text.h"
#include "repository.h"
+typedef enum {
+ GIT_CRLF_UNDEFINED,
+ GIT_CRLF_BINARY,
+ GIT_CRLF_TEXT,
+ GIT_CRLF_TEXT_INPUT,
+ GIT_CRLF_TEXT_CRLF,
+ GIT_CRLF_AUTO,
+ GIT_CRLF_AUTO_INPUT,
+ GIT_CRLF_AUTO_CRLF,
+} git_crlf_t;
+
struct crlf_attrs {
- int crlf_action;
- int eol;
+ int attr_action; /* the .gitattributes setting */
+ int crlf_action; /* the core.autocrlf setting */
+
int auto_crlf;
int safe_crlf;
+ git_cvar_value core_eol;
};
struct crlf_filter {
git_filter f;
};
-static int check_crlf(const char *value)
+static git_crlf_t check_crlf(const char *value)
{
if (GIT_ATTR_TRUE(value))
return GIT_CRLF_TEXT;
-
- if (GIT_ATTR_FALSE(value))
+ else if (GIT_ATTR_FALSE(value))
return GIT_CRLF_BINARY;
-
- if (GIT_ATTR_UNSPECIFIED(value))
- return GIT_CRLF_GUESS;
-
- if (strcmp(value, "input") == 0)
- return GIT_CRLF_INPUT;
-
- if (strcmp(value, "auto") == 0)
+ else if (GIT_ATTR_UNSPECIFIED(value))
+ ;
+ else if (strcmp(value, "input") == 0)
+ return GIT_CRLF_TEXT_INPUT;
+ else if (strcmp(value, "auto") == 0)
return GIT_CRLF_AUTO;
- return GIT_CRLF_GUESS;
+ return GIT_CRLF_UNDEFINED;
}
-static int check_eol(const char *value)
+static git_cvar_value check_eol(const char *value)
{
if (GIT_ATTR_UNSPECIFIED(value))
- return GIT_EOL_UNSET;
-
- if (strcmp(value, "lf") == 0)
+ ;
+ else if (strcmp(value, "lf") == 0)
return GIT_EOL_LF;
-
- if (strcmp(value, "crlf") == 0)
+ else if (strcmp(value, "crlf") == 0)
return GIT_EOL_CRLF;
return GIT_EOL_UNSET;
}
-static int crlf_input_action(struct crlf_attrs *ca)
-{
- if (ca->crlf_action == GIT_CRLF_BINARY)
- return GIT_CRLF_BINARY;
-
- if (ca->eol == GIT_EOL_LF)
- return GIT_CRLF_INPUT;
-
- if (ca->crlf_action == GIT_CRLF_AUTO)
- return GIT_CRLF_AUTO;
-
- if (ca->eol == GIT_EOL_CRLF)
- return GIT_CRLF_CRLF;
-
- return ca->crlf_action;
-}
-
static int has_cr_in_index(const git_filter_source *src)
{
git_repository *repo = git_filter_source_repo(src);
@@ -122,147 +112,168 @@ static int has_cr_in_index(const git_filter_source *src)
return found_cr;
}
-static int crlf_apply_to_odb(
- struct crlf_attrs *ca,
- git_buf *to,
- const git_buf *from,
- const git_filter_source *src)
+static int text_eol_is_crlf(struct crlf_attrs *ca)
{
- /* Empty file? Nothing to do */
- if (!git_buf_len(from))
+ if (ca->auto_crlf == GIT_AUTO_CRLF_TRUE)
+ return 1;
+ else if (ca->auto_crlf == GIT_AUTO_CRLF_INPUT)
return 0;
- /* Heuristics to see if we can skip the conversion.
- * Straight from Core Git.
- */
- if (ca->crlf_action == GIT_CRLF_AUTO || ca->crlf_action == GIT_CRLF_GUESS) {
- git_buf_text_stats stats;
+ if (ca->core_eol == GIT_EOL_CRLF)
+ return 1;
+ if (ca->core_eol == GIT_EOL_UNSET && GIT_EOL_NATIVE == GIT_EOL_CRLF)
+ return 1;
- /* Check heuristics for binary vs text - returns true if binary */
- if (git_buf_text_gather_stats(&stats, from, false))
- return GIT_PASSTHROUGH;
+ return 0;
+}
- /* If there are no CR characters to filter out, then just pass */
- if (!stats.cr)
- return GIT_PASSTHROUGH;
+static git_cvar_value output_eol(struct crlf_attrs *ca)
+{
+ switch (ca->crlf_action) {
+ case GIT_CRLF_BINARY:
+ return GIT_EOL_UNSET;
+ case GIT_CRLF_TEXT_CRLF:
+ return GIT_EOL_CRLF;
+ case GIT_CRLF_TEXT_INPUT:
+ return GIT_EOL_LF;
+ case GIT_CRLF_UNDEFINED:
+ case GIT_CRLF_AUTO_CRLF:
+ return GIT_EOL_CRLF;
+ case GIT_CRLF_AUTO_INPUT:
+ return GIT_EOL_LF;
+ case GIT_CRLF_TEXT:
+ case GIT_CRLF_AUTO:
+ return text_eol_is_crlf(ca) ? GIT_EOL_CRLF : GIT_EOL_LF;
+ }
+
+ /* TODO: warn when available */
+ return ca->core_eol;
+}
+
+GIT_INLINE(int) check_safecrlf(
+ struct crlf_attrs *ca,
+ const git_filter_source *src,
+ git_buf_text_stats *stats)
+{
+ const char *filename = git_filter_source_path(src);
+
+ if (!ca->safe_crlf)
+ return 0;
+
+ if (output_eol(ca) == GIT_EOL_LF) {
+ /*
+ * CRLFs would not be restored by checkout:
+ * check if we'd remove CRLFs
+ */
+ if (stats->crlf) {
+ if (ca->safe_crlf == GIT_SAFE_CRLF_WARN) {
+ /* TODO: issue a warning when available */
+ } else {
+ if (filename && *filename)
+ giterr_set(
+ GITERR_FILTER, "CRLF would be replaced by LF in '%s'",
+ filename);
+ else
+ giterr_set(
+ GITERR_FILTER, "CRLF would be replaced by LF");
- /* If safecrlf is enabled, sanity-check the result. */
- if (stats.cr != stats.crlf || stats.lf != stats.crlf) {
- switch (ca->safe_crlf) {
- case GIT_SAFE_CRLF_FAIL:
- giterr_set(
- GITERR_FILTER, "LF would be replaced by CRLF in '%s'",
- git_filter_source_path(src));
return -1;
- case GIT_SAFE_CRLF_WARN:
- /* TODO: issue warning when warning API is available */;
- break;
- default:
- break;
}
}
-
+ } else if (output_eol(ca) == GIT_EOL_CRLF) {
/*
- * We're currently not going to even try to convert stuff
- * that has bare CR characters. Does anybody do that crazy
- * stuff?
+ * CRLFs would be added by checkout:
+ * check if we have "naked" LFs
*/
- if (stats.cr != stats.crlf)
- return GIT_PASSTHROUGH;
+ if (stats->crlf != stats->lf) {
+ if (ca->safe_crlf == GIT_SAFE_CRLF_WARN) {
+ /* TODO: issue a warning when available */
+ } else {
+ if (filename && *filename)
+ giterr_set(
+ GITERR_FILTER, "LF would be replaced by CRLF in '%s'",
+ filename);
+ else
+ giterr_set(
+ GITERR_FILTER, "LF would be replaced by CRLF");
- if (ca->crlf_action == GIT_CRLF_GUESS) {
- /*
- * If the file in the index has any CR in it, do not convert.
- * This is the new safer autocrlf handling.
- */
- if (has_cr_in_index(src))
- return GIT_PASSTHROUGH;
+ return -1;
+ }
}
-
- if (!stats.cr)
- return GIT_PASSTHROUGH;
}
- /* Actually drop the carriage returns */
- return git_buf_text_crlf_to_lf(to, from);
+ return 0;
}
-static const char *line_ending(struct crlf_attrs *ca)
+static int crlf_apply_to_odb(
+ struct crlf_attrs *ca,
+ git_buf *to,
+ const git_buf *from,
+ const git_filter_source *src)
{
- switch (ca->crlf_action) {
- case GIT_CRLF_BINARY:
- case GIT_CRLF_INPUT:
- return "\n";
+ git_buf_text_stats stats;
+ bool is_binary;
+ int error;
- case GIT_CRLF_CRLF:
- return "\r\n";
+ /* Binary attribute? Empty file? Nothing to do */
+ if (ca->crlf_action == GIT_CRLF_BINARY || !git_buf_len(from))
+ return GIT_PASSTHROUGH;
- case GIT_CRLF_GUESS:
- if (ca->auto_crlf == GIT_AUTO_CRLF_FALSE)
- return "\n";
- break;
+ is_binary = git_buf_text_gather_stats(&stats, from, false);
- case GIT_CRLF_AUTO:
- if (ca->eol == GIT_EOL_CRLF)
- return "\r\n";
- case GIT_CRLF_TEXT:
- break;
+ /* Heuristics to see if we can skip the conversion.
+ * Straight from Core Git.
+ */
+ if (ca->crlf_action == GIT_CRLF_AUTO ||
+ ca->crlf_action == GIT_CRLF_AUTO_INPUT ||
+ ca->crlf_action == GIT_CRLF_AUTO_CRLF) {
- default:
- goto line_ending_error;
+ if (is_binary)
+ return GIT_PASSTHROUGH;
+
+ /*
+ * If the file in the index has any CR in it, do not convert.
+ * This is the new safer autocrlf handling.
+ */
+ if (has_cr_in_index(src))
+ return GIT_PASSTHROUGH;
}
- if (ca->auto_crlf == GIT_AUTO_CRLF_TRUE)
- return "\r\n";
- else if (ca->auto_crlf == GIT_AUTO_CRLF_INPUT)
- return "\n";
- else if (ca->eol == GIT_EOL_UNSET)
- return GIT_EOL_NATIVE == GIT_EOL_CRLF ? "\r\n" : "\n";
- else if (ca->eol == GIT_EOL_LF)
- return "\n";
- else if (ca->eol == GIT_EOL_CRLF)
- return "\r\n";
-
-line_ending_error:
- giterr_set(GITERR_INVALID, "invalid input to line ending filter");
- return NULL;
+ if ((error = check_safecrlf(ca, src, &stats)) < 0)
+ return error;
+
+ /* If there are no CR characters to filter out, then just pass */
+ if (!stats.crlf)
+ return GIT_PASSTHROUGH;
+
+ /* Actually drop the carriage returns */
+ return git_buf_text_crlf_to_lf(to, from);
}
static int crlf_apply_to_workdir(
- struct crlf_attrs *ca, git_buf *to, const git_buf *from)
+ struct crlf_attrs *ca,
+ git_buf *to,
+ const git_buf *from)
{
git_buf_text_stats stats;
- const char *workdir_ending = NULL;
bool is_binary;
/* Empty file? Nothing to do. */
- if (git_buf_len(from) == 0)
- return 0;
-
- /* Determine proper line ending */
- workdir_ending = line_ending(ca);
- if (!workdir_ending)
- return -1;
-
- /* only LF->CRLF conversion is supported, do nothing on LF platforms */
- if (strcmp(workdir_ending, "\r\n") != 0)
+ if (git_buf_len(from) == 0 || output_eol(ca) != GIT_EOL_CRLF)
return GIT_PASSTHROUGH;
- /* If there are no LFs, or all LFs are part of a CRLF, nothing to do */
is_binary = git_buf_text_gather_stats(&stats, from, false);
+ /* If there are no LFs, or all LFs are part of a CRLF, nothing to do */
if (stats.lf == 0 || stats.lf == stats.crlf)
return GIT_PASSTHROUGH;
if (ca->crlf_action == GIT_CRLF_AUTO ||
- ca->crlf_action == GIT_CRLF_GUESS) {
+ ca->crlf_action == GIT_CRLF_AUTO_INPUT ||
+ ca->crlf_action == GIT_CRLF_AUTO_CRLF) {
/* If we have any existing CR or CRLF line endings, do nothing */
- if (stats.cr > 0 && stats.crlf > 0)
- return GIT_PASSTHROUGH;
-
- /* If we have bare CR characters, do nothing */
- if (stats.cr != stats.crlf)
+ if (stats.cr > 0)
return GIT_PASSTHROUGH;
/* Don't filter binary files */
@@ -273,69 +284,80 @@ static int crlf_apply_to_workdir(
return git_buf_text_lf_to_crlf(to, from);
}
-static int crlf_check(
- git_filter *self,
- void **payload, /* points to NULL ptr on entry, may be set */
- const git_filter_source *src,
- const char **attr_values)
+static int convert_attrs(
+ struct crlf_attrs *ca,
+ const char **attr_values,
+ const git_filter_source *src)
{
int error;
- struct crlf_attrs ca;
- GIT_UNUSED(self);
+ memset(ca, 0, sizeof(struct crlf_attrs));
+
+ if ((error = git_repository__cvar(&ca->auto_crlf,
+ git_filter_source_repo(src), GIT_CVAR_AUTO_CRLF)) < 0 ||
+ (error = git_repository__cvar(&ca->safe_crlf,
+ git_filter_source_repo(src), GIT_CVAR_SAFE_CRLF)) < 0 ||
+ (error = git_repository__cvar(&ca->core_eol,
+ git_filter_source_repo(src), GIT_CVAR_EOL)) < 0)
+ return error;
+
+ /* downgrade FAIL to WARN if ALLOW_UNSAFE option is used */
+ if ((git_filter_source_flags(src) & GIT_FILTER_ALLOW_UNSAFE) &&
+ ca->safe_crlf == GIT_SAFE_CRLF_FAIL)
+ ca->safe_crlf = GIT_SAFE_CRLF_WARN;
+
+ if (attr_values) {
+ /* load the text attribute */
+ ca->crlf_action = check_crlf(attr_values[2]); /* text */
+
+ if (ca->crlf_action == GIT_CRLF_UNDEFINED)
+ ca->crlf_action = check_crlf(attr_values[0]); /* crlf */
+
+ if (ca->crlf_action != GIT_CRLF_BINARY) {
+ /* load the eol attribute */
+ int eol_attr = check_eol(attr_values[1]);
+
+ if (ca->crlf_action == GIT_CRLF_AUTO && eol_attr == GIT_EOL_LF)
+ ca->crlf_action = GIT_CRLF_AUTO_INPUT;
+ else if (ca->crlf_action == GIT_CRLF_AUTO && eol_attr == GIT_EOL_CRLF)
+ ca->crlf_action = GIT_CRLF_AUTO_CRLF;
+ else if (eol_attr == GIT_EOL_LF)
+ ca->crlf_action = GIT_CRLF_TEXT_INPUT;
+ else if (eol_attr == GIT_EOL_CRLF)
+ ca->crlf_action = GIT_CRLF_TEXT_CRLF;
+ }
- if (!attr_values) {
- ca.crlf_action = GIT_CRLF_GUESS;
- ca.eol = GIT_EOL_UNSET;
+ ca->attr_action = ca->crlf_action;
} else {
- ca.crlf_action = check_crlf(attr_values[2]); /* text */
- if (ca.crlf_action == GIT_CRLF_GUESS)
- ca.crlf_action = check_crlf(attr_values[0]); /* clrf */
- ca.eol = check_eol(attr_values[1]); /* eol */
+ ca->crlf_action = GIT_CRLF_UNDEFINED;
}
- ca.auto_crlf = GIT_AUTO_CRLF_DEFAULT;
- ca.safe_crlf = GIT_SAFE_CRLF_DEFAULT;
- /*
- * Use the core Git logic to see if we should perform CRLF for this file
- * based on its attributes & the value of `core.autocrlf`
- */
- ca.crlf_action = crlf_input_action(&ca);
+ if (ca->crlf_action == GIT_CRLF_TEXT)
+ ca->crlf_action = text_eol_is_crlf(ca) ? GIT_CRLF_TEXT_CRLF : GIT_CRLF_TEXT_INPUT;
+ if (ca->crlf_action == GIT_CRLF_UNDEFINED && ca->auto_crlf == GIT_AUTO_CRLF_FALSE)
+ ca->crlf_action = GIT_CRLF_BINARY;
+ if (ca->crlf_action == GIT_CRLF_UNDEFINED && ca->auto_crlf == GIT_AUTO_CRLF_TRUE)
+ ca->crlf_action = GIT_CRLF_AUTO_CRLF;
+ if (ca->crlf_action == GIT_CRLF_UNDEFINED && ca->auto_crlf == GIT_AUTO_CRLF_INPUT)
+ ca->crlf_action = GIT_CRLF_AUTO_INPUT;
- if (ca.crlf_action == GIT_CRLF_BINARY)
- return GIT_PASSTHROUGH;
-
- if (ca.crlf_action == GIT_CRLF_GUESS ||
- ((ca.crlf_action == GIT_CRLF_AUTO ||
- ca.crlf_action == GIT_CRLF_TEXT) &&
- git_filter_source_mode(src) == GIT_FILTER_SMUDGE)) {
-
- error = git_repository__cvar(
- &ca.auto_crlf, git_filter_source_repo(src), GIT_CVAR_AUTO_CRLF);
- if (error < 0)
- return error;
+ return 0;
+}
- if (ca.crlf_action == GIT_CRLF_GUESS &&
- ca.auto_crlf == GIT_AUTO_CRLF_FALSE)
- return GIT_PASSTHROUGH;
+static int crlf_check(
+ git_filter *self,
+ void **payload, /* points to NULL ptr on entry, may be set */
+ const git_filter_source *src,
+ const char **attr_values)
+{
+ struct crlf_attrs ca;
- if (ca.auto_crlf == GIT_AUTO_CRLF_INPUT &&
- ca.eol != GIT_EOL_CRLF &&
- git_filter_source_mode(src) == GIT_FILTER_SMUDGE)
- return GIT_PASSTHROUGH;
- }
+ GIT_UNUSED(self);
- if (git_filter_source_mode(src) == GIT_FILTER_CLEAN) {
- error = git_repository__cvar(
- &ca.safe_crlf, git_filter_source_repo(src), GIT_CVAR_SAFE_CRLF);
- if (error < 0)
- return error;
+ convert_attrs(&ca, attr_values, src);
- /* downgrade FAIL to WARN if ALLOW_UNSAFE option is used */
- if ((git_filter_source_flags(src) & GIT_FILTER_ALLOW_UNSAFE) &&
- ca.safe_crlf == GIT_SAFE_CRLF_FAIL)
- ca.safe_crlf = GIT_SAFE_CRLF_WARN;
- }
+ if (ca.crlf_action == GIT_CRLF_BINARY)
+ return GIT_PASSTHROUGH;
*payload = git__malloc(sizeof(ca));
GITERR_CHECK_ALLOC(*payload);
@@ -345,15 +367,16 @@ static int crlf_check(
}
static int crlf_apply(
- git_filter *self,
- void **payload, /* may be read and/or set */
- git_buf *to,
+ git_filter *self,
+ void **payload, /* may be read and/or set */
+ git_buf *to,
const git_buf *from,
const git_filter_source *src)
{
/* initialize payload in case `check` was bypassed */
if (!*payload) {
int error = crlf_check(self, payload, src, NULL);
+
if (error < 0)
return error;
}
diff --git a/src/filter.h b/src/filter.h
index b1c403ba9..34081fb4a 100644
--- a/src/filter.h
+++ b/src/filter.h
@@ -15,16 +15,6 @@
/* Amount of file to examine for NUL byte when checking binary-ness */
#define GIT_FILTER_BYTES_TO_CHECK_NUL 8000
-/* Possible CRLF values */
-typedef enum {
- GIT_CRLF_GUESS = -1,
- GIT_CRLF_BINARY = 0,
- GIT_CRLF_TEXT,
- GIT_CRLF_INPUT,
- GIT_CRLF_CRLF,
- GIT_CRLF_AUTO,
-} git_crlf_t;
-
typedef struct {
git_attr_session *attr_session;
git_buf *temp_buf;