commit 5dec57616c9ae8a4ed8dfe36a48572f453be7b13
parent 6d10b98b51a77588609429ea77fa5ca396ced0a9
Author: Drew DeVault <sir@cmpwn.com>
Date: Fri, 2 Apr 2021 17:42:08 -0400
hare::module: prune down less-specific candidates
Diffstat:
4 files changed, 100 insertions(+), 23 deletions(-)
diff --git a/hare/module/scan.ha b/hare/module/scan.ha
@@ -13,6 +13,7 @@ use slice;
use sort;
use strings;
use strio;
+use fmt;
// Scans the files in a directory for eligible build inputs and returns a
// [version] which includes all applicable files and their dependencies.
@@ -58,8 +59,8 @@ export fn scan(ctx: *context, path: str) (version | error) = {
return ver;
};
-// Given a name and whether or not it represents a directory, parses it into the
-// basename, extension, and tag set.
+// Given a file or directory name, parses it into the basename, extension, and
+// tag set.
fn parse_name(name: str) (str, str, []tag) = {
let ext = path::extension(name);
let base = ext.0, ext = ext.1;
@@ -122,14 +123,40 @@ fn scan_directory(
};
};
+ // Sorted to keep the hash consistent
sort::sort(dirs, size(str), &strcmp);
sort::sort(files, size(str), &strcmp);
- // TODO: filter out less specific tag sets for a given basename
+ // Tuple of is_directory, basename, tags, and path to a candidate input.
+ let inputs: [](bool, str, []tag, str) = [];
+ defer for (let i = 0z; i < len(inputs); i += 1) {
+ // For file paths, these are assigned to the input, which
+ // assumes ownership over them.
+ if (inputs[i].0) {
+ free(inputs[i].1);
+ tags_free(inputs[i].2);
+ free(inputs[i].3);
+ };
+ };
+
+ // For a given basename, only the most specific path (i.e. with the most
+ // tags) is used.
+ //
+ // foo.ha
+ // foo+linux.ha
+ // foo+linux+x86_64/
+ // bar.ha
+ // baz.ha
+ //
+ // In this case, foo+linux+x86_64 is the most specific, and so its used
+ // as the build input and the other two files are discarded.
+ //
+ // TODO: Improve the documentation which describes this algorithm
+
for (let i = 0z; i < len(dirs); i += 1) {
let name = dirs[i];
- let tags = parse_name(name).2;
- defer tags_free(tags);
+ let parsed = parse_name(name);
+ let base = parsed.0, tags = parsed.2;
let d = strings::toutf8(name);
if (len(d) == 0 || (
@@ -141,9 +168,25 @@ fn scan_directory(
continue;
};
- let p = path::join(path, name);
- let iter = fs::iter(ctx.fs, p)?;
- scan_directory(ctx, ver, sha, p, iter)?;
+ let path = path::join(path, name);
+ let tuple = (true, strings::dup(base), tags, path);
+ let superceeded = false;
+ for (let j = 0z; j < len(inputs); j += 1) {
+ if (inputs[j].1 != base) {
+ continue;
+ };
+ if (len(inputs[j].2) < len(tags)) {
+ free(inputs[j].1);
+ tags_free(inputs[j].2);
+ free(inputs[j].3);
+ inputs[j] = tuple;
+ superceeded = true;
+ break;
+ };
+ };
+ if (!superceeded) {
+ append(inputs, tuple);
+ };
};
for (let i = 0z; i < len(files); i += 1) {
@@ -164,19 +207,53 @@ fn scan_directory(
continue;
};
- let p = path::join(path, name);
- let st = fs::stat(ctx.fs, p)?;
- let in = input {
- path = fs::resolve(ctx.fs, p),
- stat = st,
- ft = type_for_ext(name) as filetype,
- hash = scan_file(ctx, p, &ver.depends)?,
- basename = base,
- tags = tags,
- ...
+ let path = path::join(path, name);
+ let tuple = (false, base, tags, path);
+ let superceeded = false;
+ for (let j = 0z; j < len(inputs); j += 1) {
+ if (inputs[j].1 != base) {
+ continue;
+ };
+ if (len(inputs[j].2) < len(tags)) {
+ // We are more specific
+ free(inputs[j].1);
+ tags_free(inputs[j].2);
+ free(inputs[j].3);
+ inputs[j] = tuple;
+ superceeded = true;
+ break;
+ } else if (len(inputs[j].2) > len(tags)) {
+ // They are more specific
+ superceeded = true;
+ break;
+ } else {
+ abort(); // TODO
+ };
+ };
+ if (!superceeded) {
+ append(inputs, tuple);
+ };
+ };
+
+ for (let i = 0z; i < len(inputs); i += 1) {
+ let isdir = inputs[i].0, path = inputs[i].3;
+ if (isdir) {
+ let iter = fs::iter(ctx.fs, path)?;
+ scan_directory(ctx, ver, sha, path, iter)?;
+ } else {
+ let st = fs::stat(ctx.fs, path)?;
+ let in = input {
+ path = fs::resolve(ctx.fs, path),
+ stat = st,
+ ft = type_for_ext(path) as filetype,
+ hash = scan_file(ctx, path, &ver.depends)?,
+ basename = inputs[i].1,
+ tags = inputs[i].2,
+ ...
+ };
+ append(ver.inputs, in);
+ hash::write(sha, in.hash);
};
- append(ver.inputs, in);
- hash::write(sha, in.hash);
};
};
diff --git a/rt/+linux/start.ha b/rt/+linux/platformstart.ha
diff --git a/scripts/gen-stdlib b/scripts/gen-stdlib
@@ -21,7 +21,7 @@ gensrcs_rt() {
'$(PLATFORM)/errno.ha' \
'$(PLATFORM)/types.ha' \
'$(PLATFORM)/segmalloc.ha' \
- '$(PLATFORM)/start.ha' \
+ '$(PLATFORM)/platformstart.ha' \
'$(PLATFORM)/$(ARCH).ha' \
'$(PLATFORM)/syscallno$(ARCH).ha' \
'$(PLATFORM)/syscalls.ha' \
diff --git a/stdlib.mk b/stdlib.mk
@@ -8,7 +8,7 @@ stdlib_rt_srcs= \
$(STDLIB)/rt/$(PLATFORM)/errno.ha \
$(STDLIB)/rt/$(PLATFORM)/types.ha \
$(STDLIB)/rt/$(PLATFORM)/segmalloc.ha \
- $(STDLIB)/rt/$(PLATFORM)/start.ha \
+ $(STDLIB)/rt/$(PLATFORM)/platformstart.ha \
$(STDLIB)/rt/$(PLATFORM)/$(ARCH).ha \
$(STDLIB)/rt/$(PLATFORM)/syscallno$(ARCH).ha \
$(STDLIB)/rt/$(PLATFORM)/syscalls.ha \
@@ -768,7 +768,7 @@ testlib_rt_srcs= \
$(STDLIB)/rt/$(PLATFORM)/errno.ha \
$(STDLIB)/rt/$(PLATFORM)/types.ha \
$(STDLIB)/rt/$(PLATFORM)/segmalloc.ha \
- $(STDLIB)/rt/$(PLATFORM)/start.ha \
+ $(STDLIB)/rt/$(PLATFORM)/platformstart.ha \
$(STDLIB)/rt/$(PLATFORM)/$(ARCH).ha \
$(STDLIB)/rt/$(PLATFORM)/syscallno$(ARCH).ha \
$(STDLIB)/rt/$(PLATFORM)/syscalls.ha \
