RFE: implement fast mode for getting EVR

[ignatenkobrain]

rpmspec is quite slow because it needs to parse whole spec and so on. Would be nice if there would be some flag to turn off %include, parsing of all sections and just give me stupid EVR. It won't check spec for correctness and such, but it would be very fast.

Just to give some background, I'm writing tool which is generating RPM changelogs out of git history and rpmspec is slowest part in there.

[ignatenkobrain]

@n3npq believe me, it is slowest part in my tool ;)

287.25user 152.95system 2:11.94elapsed 333%CPU (0avgtext+0avgdata 68724maxresident)k
13648inputs+0outputs (53major+30349595minor)pagefaults 0swaps

diff --git a/src/main.rs b/src/main.rs
index 4b0f2ff..9306517 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -79,6 +79,7 @@ fn run(opt: &Opt) -> Result<(), Error> {
             let atime = author.when();
             let datetime =
                 FixedOffset::east(atime.offset_minutes() * 3600 / 60).timestamp(atime.seconds(), 0);
+            /*
             let output = Command::new("rpmspec")
                 .args(&[
                     "--srpm",
@@ -92,6 +93,7 @@ fn run(opt: &Opt) -> Result<(), Error> {
                     workdir.join(&spec).to_str().unwrap(),
                 ])
                 .output()?;
+            */
 
             let mut chlog_header = format!(
                 "* {} {} <{}>",
@@ -99,11 +101,13 @@ fn run(opt: &Opt) -> Result<(), Error> {
                 author.name().unwrap_or("Nobody"),
                 author.email().unwrap_or("nobody@fedoraproject.org"),
             );
+            /*
             if output.status.success() {
                 chlog_header.push_str(&format!(" - {}", String::from_utf8_lossy(&output.stdout)));
             } else {
                 eprintln!("{}", String::from_utf8_lossy(&output.stderr));
             }
+            */
             let chlog_entry = format!("- {}", commit.summary().unwrap());
 
             Ok(Some(format!("{}\n{}", chlog_header, chlog_entry)))

13.41user 2.05system 0:03.20elapsed 482%CPU (0avgtext+0avgdata 66184maxresident)k
0inputs+0outputs (0major+495394minor)pagefaults 0swaps

[ignatenkobrain]

@n3npq it's not single spec.. Its 4000+ different specs.

What I wrote there if you run rpmspec, then tool is taking 2 minutes to do the job. If you remove execution of rpmspec, then tool takes 3 seconds.

[ignatenkobrain]

Since your program appears to be multithreaded, try splitting the 4000 specfile queries across 10-20 worker threads.

That's exactly what it does.

[ignatenkobrain]

If I would be querying them sequentially, then it would take hour or so

[ffesti]

yup, wontfix.